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 GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,balanced,0.05393599967161814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,balanced,0.05462933580080668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,balanced,0.06058133145173391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,balanced,0.06465066472689311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,balanced,0.07418133318424225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,balanced,0.07366399963696797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,balanced,0.07851733267307281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.1153600017229716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.12786666552225748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.15640532970428467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.18497065703074136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.23938665787378946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.29705599943796795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.4081439971923828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.5231893459955851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09379839897155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09674879908561707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.09691519737243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.09726079702377319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.09678720235824585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.09845119714736938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.11121920347213746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.11416959762573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.12624640464782716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.13194880485534669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.13896960020065308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.1684671998023987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.1898687958717346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.24557440280914306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.28758399486541747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.40103678703308104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.5053696155548095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7252416133880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.9273856163024903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.380742359161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.7945344924926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,power_law_1.2,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,power_law_1.2,0.060108798742294314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,power_law_1.2,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,power_law_1.2,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,power_law_1.2,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,power_law_1.2,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,power_law_1.2,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,power_law_1.2,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,power_law_1.2,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,power_law_1.2,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,power_law_1.2,0.09013760089874268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.2,0.12329599857330323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.2,0.14526720046997071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.2,0.18260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.2,0.22556159496307374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.2,0.36062719821929934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.2,0.43080959320068357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.2,0.7095039844512939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.2,0.9149120330810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,balanced,0.03506666670242945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.07352533439795177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.09400000174840291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.10846400260925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.14639467000961304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.1815626621246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,balanced,0.06007466713587443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,balanced,0.08025600016117096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,balanced,0.10639466842015584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,balanced,0.1557813286781311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,balanced,0.2592479983965556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,balanced,0.45982933044433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,balanced,0.46377066771189374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,balanced,0.4617439905802409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,balanced,0.4732960065205892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,balanced,0.4787840048472087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,balanced,0.47439467906951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,balanced,0.4829920132954915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,balanced,0.4885760148366292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,balanced,0.4967840115229289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,balanced,0.5035093228022257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,balanced,0.5164639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,balanced,0.5369706551233927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,balanced,0.5772000153859457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,balanced,0.6064800024032593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,balanced,0.6919466654459635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,balanced,0.7638826370239258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,balanced,0.9762933254241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,balanced,1.1035466988881428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,balanced,1.5355413754781086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,balanced,1.8009227116902669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,balanced,2.6070399284362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,balanced,3.2954346338907876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,power_law_1.2,0.07863680124282837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,power_law_1.2,0.11157759428024291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,power_law_1.2,0.1482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,power_law_1.2,0.23237121105194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,power_law_1.2,0.31111679077148435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,power_law_1.2,0.32429440021514894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,power_law_1.2,0.3321984052658081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,power_law_1.2,0.34771199226379396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,power_law_1.2,0.3543040037155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,power_law_1.2,0.3652224063873291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,power_law_1.2,0.3770944118499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,power_law_1.2,0.38417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,power_law_1.2,0.4010623931884766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,power_law_1.2,0.41527681350708007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,power_law_1.2,0.43445119857788084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,power_law_1.2,0.48280959129333495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.2,0.5216447830200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.2,0.6134655952453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.2,0.7006976127624511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.2,0.8652159690856933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.2,1.0152959823608398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.2,1.3326335906982423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.2,1.6091840744018555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.2,2.222483253479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.2,2.8161600112915037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.24428799152374267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.3008960008621216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.44872322082519533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.6074687957763671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.8217791557312012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.1480511665344237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.2427007675170898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.2570303916931151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.2102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.3780735969543456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.400928020477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.418278408050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.5084287643432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.4556287765502929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,power_law_1.2,1.5443519592285155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,power_law_1.2,1.5846272468566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,power_law_1.2,1.748953628540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,power_law_1.2,1.8261184692382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,power_law_1.2,2.0394559860229493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,power_law_1.2,2.000614356994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,power_law_1.2,2.212403106689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,power_law_1.2,2.44005126953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,power_law_1.2,2.822528076171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,power_law_1.2,3.4570686340332033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,power_law_1.2,4.270284652709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,power_law_1.2,5.219148635864258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,balanced,0.04082666585842768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,balanced,0.08555733164151509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,balanced,0.08746133248011272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,balanced,0.08826667070388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,balanced,0.09252267082532246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,balanced,0.09284800291061401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,balanced,0.09657599528630574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,balanced,0.09674132863680522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,balanced,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,balanced,0.10942932963371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,balanced,0.11472533146540324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,balanced,0.12000000476837158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,balanced,0.12728533148765564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,balanced,0.1474346617857615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,balanced,0.1644373337427775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,balanced,0.20323733488718668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,balanced,0.23189866542816162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,balanced,0.3278026580810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,balanced,0.4013599952061971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,balanced,0.5832159916559855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,balanced,0.7499573230743408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,balanced,1.091312011082967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,balanced,1.4298399289449055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,power_law_1.01,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,power_law_1.01,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,power_law_1.01,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,power_law_1.01,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,power_law_1.01,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,power_law_1.01,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,power_law_1.01,0.07296640276908875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,power_law_1.01,0.07946879863739013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,power_law_1.01,0.08165119886398316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,power_law_1.01,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,power_law_1.01,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,power_law_1.01,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,power_law_1.01,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,power_law_1.01,0.11944960355758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,power_law_1.01,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,power_law_1.01,0.15475840568542482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,power_law_1.01,0.17913600206375122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,power_law_1.01,0.23639678955078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,power_law_1.01,0.30268800258636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,power_law_1.01,0.40135040283203127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,power_law_1.01,0.4941823959350586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,power_law_1.01,0.6826367855072022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,power_law_1.01,0.8972415924072266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.048332801461219786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.13696639537811278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.1415168046951294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.14831360578536987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.14792959690093993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.15054719448089598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.15055999755859376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.16147840023040771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.16195839643478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.16795519590377808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.17779200077056884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.18437119722366332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.1927296042442322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.22442240715026857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.2430016040802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.29750399589538573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.35289599895477297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.4329855918884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.5398464202880859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.6839615821838378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,0.8856255531311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,1.2650815963745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.18676480054855346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.23610880374908447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.20902400016784667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.28384640216827395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.27613439559936526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.3106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.28752000331878663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.29959681034088137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.3084671974182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.312940788269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.33777918815612795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.3226367950439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.35715839862823484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.3806335926055908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.39492480754852294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.45844478607177735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.49102082252502444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.6193215847015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,power_law_1.2,0.6615039825439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,power_law_1.2,0.9020095825195312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,power_law_1.2,1.0355711936950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,power_law_1.2,1.360262393951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,power_law_1.2,1.7254207611083985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,1.5044032096862794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,power_law_1.01,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,power_law_1.01,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,power_law_1.01,0.048582398891448976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,power_law_1.01,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,power_law_1.01,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,power_law_1.01,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,power_law_1.01,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,power_law_1.01,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.01,0.08486400246620178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.01,0.09277439713478089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.01,0.11547520160675048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.01,0.1876799941062927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.01,0.23214080333709716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.01,0.3226111888885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.01,0.4081408023834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,balanced,0.029450667401154835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,balanced,0.03323733309904734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,balanced,0.03309333324432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,balanced,0.033071999748547874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,balanced,0.03796799977620443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,balanced,0.06379733482996623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,balanced,0.09173867106437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,balanced,0.10726400216420491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,balanced,0.14511467019716898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,balanced,0.17589332660039267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,power_law_1.2,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,power_law_1.2,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,power_law_1.2,0.033939200639724734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,power_law_1.2,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,power_law_1.2,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,power_law_1.2,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,power_law_1.2,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,power_law_1.2,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,power_law_1.2,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.2,0.11630719900131226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.2,0.15886720418930053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.2,0.20476799011230468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.034815999865531924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.08051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.09274880290031433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.17019519805908204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.22887680530548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.2669055938720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.3757823944091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.4836991786956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.6575424194335937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,power_law_1.01,0.9179903984069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.2,0.26998400688171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.2,0.3538496017456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.2,0.5148159980773925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.2,0.7759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.2,1.2015423774719238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.09530879855155945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.10247679948806762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.11263359785079956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.11651840209960937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.11963520050048829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.1369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.150655996799469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.1749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.2116607904434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.2441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.31422080993652346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.3899519920349121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.5505663871765136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.6896383762359619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,1.0035200119018555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.2553024291992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,balanced,0.02940266579389572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,balanced,0.03384533276160558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,balanced,0.03257599969704946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,balanced,0.03322133421897888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,balanced,0.042463997999827065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,balanced,0.0793333351612091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,balanced,0.11307733257611592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,balanced,0.1421280006567637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,balanced,0.1925813357035319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,balanced,0.23346134026845297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.2,0.05489919781684875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.2,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.2,0.029951998591423036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.2,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.2,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.2,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.2,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.2,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.2,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.2,0.07183359861373902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.2,0.11556479930877686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.2,0.13544319868087767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.2,0.17404160499572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.2,0.231712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.2,0.32709760665893556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,balanced,0.13733333349227905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,balanced,0.23477866252263388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,balanced,0.44250667095184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,balanced,0.443178653717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,balanced,0.43728001912434894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,balanced,0.440720001856486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,balanced,0.4410933256149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,balanced,0.44449599583943683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,balanced,0.44396265347798664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,balanced,0.44790399074554443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,balanced,0.4528426726659139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,balanced,0.4537706772486369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,balanced,0.4596480131149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,balanced,0.4779733419418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,balanced,0.4835840066274007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,balanced,0.4901226758956909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,balanced,0.5741066535313925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,balanced,0.5605973402659098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,balanced,0.6082826852798462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,balanced,0.7858026822408041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,balanced,0.8767039775848389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,balanced,1.2225120067596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,balanced,1.4824320475260417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,balanced,2.1834774017333984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,balanced,2.8701279958089194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,balanced,4.255242665608724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,balanced,5.607674916585286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,power_law_1.01,0.025920000672340394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,power_law_1.01,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,power_law_1.01,0.025260800123214723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,power_law_1.01,0.02770560085773468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,power_law_1.01,0.0290367990732193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,power_law_1.01,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,power_law_1.01,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,power_law_1.01,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,power_law_1.01,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,power_law_1.01,0.09909759759902954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,power_law_1.01,0.12728960514068605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,power_law_1.01,0.2029184103012085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,power_law_1.01,0.2647871971130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,power_law_1.01,0.3605823993682861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,power_law_1.01,0.4401408195495605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,power_law_1.2,0.03075839877128601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,power_law_1.2,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,power_law_1.2,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,power_law_1.2,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,power_law_1.2,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,power_law_1.2,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,power_law_1.2,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,power_law_1.2,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,power_law_1.2,0.1132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,power_law_1.2,0.15661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,power_law_1.2,0.1827072024345398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,power_law_1.2,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,balanced,0.040933333337306976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.01,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,balanced,0.05037866532802582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,balanced,0.051701332132021584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.01,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.01,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,balanced,0.07891199986139934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,balanced,0.0869706670443217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.01,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,balanced,0.10123733679453532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,balanced,0.11754133303960164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,balanced,0.15896000464757284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,balanced,0.19131199518839517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,balanced,0.26360533634821576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,balanced,0.33371198177337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.01,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.01,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.01,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.01,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.01,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,power_law_1.2,0.3024447917938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.01,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.01,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.01,0.10640000104904175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.01,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.01,0.14009599685668944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.01,0.18592000007629395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.01,0.2507328033447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.01,0.2936896085739136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,power_law_1.2,0.40480642318725585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,power_law_1.2,0.5961535930633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,balanced,0.06617600222428639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,balanced,0.08752533793449402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,balanced,0.1234879990418752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,balanced,0.19481066862742105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,balanced,0.3261760075887044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,balanced,0.4535413185755412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,balanced,0.4572106599807739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,balanced,0.4573226769765218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,balanced,0.4617439905802409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,balanced,0.4618719816207886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,balanced,0.4652640024820964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,balanced,0.4715199867884318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,balanced,0.47255468368530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,balanced,0.4822880029678345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,balanced,0.4934026797612508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,balanced,0.5008960167566935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,balanced,0.5162773529688517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,balanced,0.6312693357467651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,balanced,0.5791200002034506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,balanced,0.7306239604949951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,balanced,0.7177813053131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,balanced,1.0324479738871257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,balanced,0.9787680308024088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,balanced,1.3899572690327961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,balanced,1.5531039237976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,balanced,2.4347732861836753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,balanced,2.761749267578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,balanced,0.049626668294270836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,balanced,0.056320001681645714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,balanced,0.09026666482289632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,balanced,0.09236266215642293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,balanced,0.09498666723569234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,balanced,0.09267200032869975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,balanced,0.09503466884295146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,balanced,0.09410132964452107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,balanced,0.09496000409126282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,balanced,0.09658666451772054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,balanced,0.0962720016638438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,balanced,0.10337600111961365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,balanced,0.1053653359413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,balanced,0.10763200124104817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,balanced,0.11133866508801778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,balanced,0.11687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,balanced,0.13032000263532004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,balanced,0.13892799615859985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,balanced,0.1644159952799479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,balanced,0.18549867471059164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,balanced,0.2424479921658834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,balanced,0.2887466748555501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,balanced,0.40426135063171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,balanced,0.5201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,balanced,0.05431999762852987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,balanced,0.05473599831263224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,balanced,0.06214400132497152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,balanced,0.06135466694831848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,balanced,0.07717866698900859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,balanced,0.07668800155321757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,balanced,0.10257599751154582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,balanced,0.11833066741625468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,balanced,0.1511626640955607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,balanced,0.17212265729904175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,balanced,0.22713599602381387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,balanced,0.27157866954803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,balanced,0.3737866481145223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,balanced,0.4845920006434123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.01,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.01,0.10471680164337158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.01,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.01,0.0961408019065857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.01,0.102457594871521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.01,0.09722239971160888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.01,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.01,0.10800000429153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.01,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.01,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.01,0.11789439916610718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.01,0.11889280080795288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.01,0.1301632046699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.01,0.1493888020515442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.01,0.17066880464553832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.01,0.2048703908920288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.01,0.22832000255584717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.01,0.2937920093536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.01,0.29285120964050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.01,0.39051520824432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.01,0.4897471904754639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.01,0.6426176071166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.01,0.8173312187194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,power_law_1.2,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,power_law_1.2,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,power_law_1.2,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,power_law_1.2,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,power_law_1.2,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,power_law_1.2,0.10308480262756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,power_law_1.2,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,power_law_1.2,0.13626879453659058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,power_law_1.2,0.17173759937286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,power_law_1.2,0.22787199020385743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,power_law_1.2,0.2809983968734741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,power_law_1.2,0.4181568145751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,power_law_1.2,0.5354879856109619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,power_law_1.2,0.7719808101654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,power_law_1.2,0.9944128036499024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,power_law_1.01,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,power_law_1.01,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,power_law_1.01,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,power_law_1.01,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,power_law_1.01,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,power_law_1.01,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,power_law_1.01,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,power_law_1.01,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,power_law_1.01,0.1160256028175354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,power_law_1.01,0.13797119855880738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,power_law_1.01,0.18561279773712158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,power_law_1.01,0.22862720489501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,power_law_1.01,0.3243072032928467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,power_law_1.01,0.4350143909454346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,power_law_1.01,0.6330239772796631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,power_law_1.01,0.8277183532714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,balanced,0.06996266543865204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,balanced,0.09975467125574748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,balanced,0.14863466223080954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,balanced,0.1480959951877594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,balanced,0.1507146656513214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,balanced,0.1511573294798533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,balanced,0.1514026621977488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,balanced,0.15294399857521057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,balanced,0.15596266587575278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,balanced,0.15595733126004538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,balanced,0.1597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,balanced,0.16749866803487143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,balanced,0.16854933897654215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,balanced,0.1754186749458313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,balanced,0.1850879987080892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,balanced,0.19649600982666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,balanced,0.22159467140833536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,balanced,0.25496532519658405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,balanced,0.3038453261057536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,balanced,0.3508693377176921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,balanced,0.449504017829895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,balanced,0.5397866566975912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,balanced,0.7408320109049479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,balanced,0.9307306607564291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,power_law_1.2,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,power_law_1.2,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,power_law_1.2,0.09503359794616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,power_law_1.2,0.11589759588241577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,power_law_1.2,0.11797759532928467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,power_law_1.2,0.1196671962738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,power_law_1.2,0.12526079416275024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,power_law_1.2,0.13267199993133544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,power_law_1.2,0.13914239406585693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,power_law_1.2,0.14303359985351563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,power_law_1.2,0.1553663969039917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,power_law_1.2,0.16722559928894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,power_law_1.2,0.17790720462799073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,power_law_1.2,0.17560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,power_law_1.2,0.20638079643249513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,power_law_1.2,0.23401598930358886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,power_law_1.2,0.2790271997451782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,power_law_1.2,0.33432960510253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,power_law_1.2,0.4443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,power_law_1.2,0.5846975803375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,power_law_1.2,0.7602752208709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,power_law_1.2,1.0131903648376466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,power_law_1.2,1.3562175750732421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,power_law_1.2,1.7992191314697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,balanced,0.047482664386431374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,balanced,0.048623998959859215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,balanced,0.07558399935563405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,balanced,0.10502933462460835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,balanced,0.15455466508865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,balanced,0.15769066413243613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,balanced,0.15646933515866598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,balanced,0.1606986622015635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,balanced,0.1600053310394287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,balanced,0.16335466504096985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,balanced,0.16504533092180887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,balanced,0.16775466998418173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,balanced,0.1696853240331014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,balanced,0.18311466773351034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,balanced,0.18620266517003378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,balanced,0.20733332633972168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,balanced,0.2203786571820577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,balanced,0.2514080007870992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,balanced,0.2839840054512024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,balanced,0.3531200091044108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,balanced,0.41568533579508465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,balanced,0.5627413193384806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,balanced,0.6831253369649252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,balanced,0.9591146310170492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,balanced,1.2148213386535645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,balanced,0.029994666576385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,balanced,0.032618666688601174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,balanced,0.0391893337170283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,balanced,0.041162667175134025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,balanced,0.04018666595220566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,balanced,0.04820266862710317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,balanced,0.08473066488901775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,balanced,0.11392533779144287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,balanced,0.1409173309803009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,balanced,0.1851146618525187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,balanced,0.23243200778961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,balanced,0.33318932851155597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,balanced,0.4150933424631755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.09177600145339966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.08222079873085023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.11965440511703491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.13030400276184081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.1555008053779602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.17161600589752196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.2191551923751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.25518720149993895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.35550079345703123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.49541759490966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.6609663963317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.8343104362487793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,balanced,0.04880533119042715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,balanced,0.04659733176231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,balanced,0.04685866832733154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,balanced,0.04966400067011515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,balanced,0.05528533458709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,balanced,0.08075200021266937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,balanced,0.107205331325531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,balanced,0.14100799957911173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,balanced,0.17035732666651407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,balanced,0.23093332846959433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,balanced,0.2940319975217183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,power_law_1.01,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,power_law_1.01,0.03362559974193573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,power_law_1.01,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,power_law_1.01,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,power_law_1.01,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,power_law_1.01,0.042483198642730716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,power_law_1.01,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,power_law_1.01,0.048127999901771544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,power_law_1.01,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,power_law_1.01,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,power_law_1.01,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,power_law_1.01,0.20401279926300048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,power_law_1.01,0.2640511989593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,power_law_1.01,0.44033279418945315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,power_law_1.01,0.5265279769897461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,power_law_1.01,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,power_law_1.01,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07627519965171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,power_law_1.01,0.08019199967384338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,power_law_1.01,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,power_law_1.01,0.09051520228385926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,power_law_1.01,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,power_law_1.01,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,power_law_1.01,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,power_law_1.01,0.13364479541778565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.14529279470443726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.17809280157089233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.19719680547714233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2609472036361694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3253376007080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.4484864234924316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.5662464141845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8035072326660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.0269887924194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.01,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.01,0.1480191946029663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.01,0.13352320194244385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.01,0.1534208059310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.01,0.17585279941558837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.01,0.18113919496536254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.01,0.1782912015914917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.01,0.17361279726028442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.01,0.17450239658355712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.01,0.17858560085296632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.01,0.18795520067214966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.01,0.19613440036773683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.01,0.19056639671325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.01,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.01,0.20691840648651122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.01,0.20590078830718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.01,0.2163072109222412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.01,0.24394240379333496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.01,0.2767872095108032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.01,0.3358911991119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.01,0.3840127944946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.01,0.51178879737854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.01,0.5626175880432129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.01,0.7452864170074462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.01,0.9023103713989258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.01,1.2289664268493652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.01,1.580953598022461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,power_law_1.2,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,power_law_1.2,0.10931199789047241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,power_law_1.2,0.13127039670944213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,power_law_1.2,0.13458559513092042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,power_law_1.2,0.12996480464935303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,power_law_1.2,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,power_law_1.2,0.14332799911499022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,power_law_1.2,0.14783999919891358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,power_law_1.2,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,power_law_1.2,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,power_law_1.2,0.16766079664230346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,power_law_1.2,0.17756160497665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,power_law_1.2,0.20473599433898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.2,0.23031680583953856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.2,0.28060159683227537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.2,0.3298304080963135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.2,0.42924160957336427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.2,0.5436992168426513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.2,0.7462592124938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.2,0.9366144180297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.2,1.4322303771972655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.2,1.829075241088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,power_law_1.2,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,power_law_1.2,0.046239998936653134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,power_law_1.2,0.04442879855632782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,power_law_1.2,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,power_law_1.2,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,power_law_1.2,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,power_law_1.2,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,power_law_1.2,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,power_law_1.2,0.04785920083522797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,power_law_1.2,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,power_law_1.2,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,power_law_1.2,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,power_law_1.2,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,power_law_1.2,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,power_law_1.2,0.12335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,power_law_1.2,0.15104000568389891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,power_law_1.2,0.22183680534362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,power_law_1.2,0.28580479621887206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,power_law_1.2,0.3855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,power_law_1.2,0.48706560134887694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,balanced,0.042624001701672874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,balanced,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,balanced,0.04444799820582072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,balanced,0.04419733087221781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,balanced,0.04225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,balanced,0.04537599782148997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,balanced,0.04773866633574168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,balanced,0.06273066500822704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,balanced,0.06479466458161671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,balanced,0.08097599943478902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,balanced,0.07718933125336964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,balanced,0.09939733147621155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,balanced,0.111135999361674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,balanced,0.14461867014567056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,balanced,0.18109333515167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,balanced,0.2530240019162496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,balanced,0.3339200019836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,balanced,0.48602668444315594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,balanced,0.6295520067214966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,power_law_1.01,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,power_law_1.01,0.07472000122070313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,power_law_1.01,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,power_law_1.01,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,power_law_1.01,0.07832959890365601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,power_law_1.01,0.08239359855651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,power_law_1.01,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,power_law_1.01,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,power_law_1.01,0.09405440092086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,power_law_1.01,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,power_law_1.01,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,power_law_1.01,0.12800639867782593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,power_law_1.01,0.14847359657287598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,power_law_1.01,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,power_law_1.01,0.19728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,power_law_1.01,0.22117118835449218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,power_law_1.01,0.3174207925796509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,power_law_1.01,0.4175104141235352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,power_law_1.01,0.5253439903259277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,power_law_1.01,0.6798463821411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.06543359756469727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.06874880194664001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.07166079878807068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.08202239871025085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.09239680171012879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.14067840576171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.1593727946281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.1981951951980591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.23031680583953856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.3222143888473511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.44381442070007326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.6021759986877442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.7608255863189697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,balanced,0.07094933092594147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,balanced,0.0727893312772115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,balanced,0.0746613343556722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,balanced,0.09489066402117412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,balanced,0.11348266402880351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,balanced,0.1497813363869985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,balanced,0.1507146656513214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,balanced,0.15307199954986572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,balanced,0.15037866433461508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,balanced,0.15385599931081137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,balanced,0.15454933047294617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,balanced,0.15582933028539023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,balanced,0.15863466262817383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,balanced,0.16132799784342447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,balanced,0.16745599110921225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,balanced,0.16932799418767294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,balanced,0.17485332489013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,balanced,0.18856000900268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,balanced,0.2006666660308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,balanced,0.2404373288154602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,balanced,0.26949334144592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,balanced,0.3415466547012329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,balanced,0.40507733821868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,balanced,0.5771733522415161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,balanced,0.70796799659729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,balanced,1.0184319814046223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,balanced,1.3402719497680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,power_law_1.01,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,power_law_1.01,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,power_law_1.01,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,power_law_1.01,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,power_law_1.01,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,power_law_1.01,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,power_law_1.01,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,power_law_1.01,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,power_law_1.01,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,power_law_1.01,0.13212159872055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,power_law_1.01,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,power_law_1.01,0.22383360862731932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,power_law_1.01,0.28257920742034914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,power_law_1.01,0.4048448085784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,power_law_1.01,0.5074944019317627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,power_law_1.2,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,power_law_1.2,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,power_law_1.2,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,power_law_1.2,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,power_law_1.2,0.06175360083580017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,power_law_1.2,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.2,0.08295040130615235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.2,0.1401535987854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.2,0.16651519536972045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.2,0.23458559513092042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.2,0.2865151882171631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.2,0.4218111991882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.2,0.5357312202453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.13866879940032958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.18097920417785646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.2095103979110718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,power_law_1.2,0.284332799911499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,power_law_1.2,0.357094407081604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,power_law_1.2,0.5661375999450684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,power_law_1.2,0.6321087837219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.12833919525146484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.13994879722595216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.1705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.16908799409866332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.1770624041557312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.17641600370407104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.16780799627304077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.16989439725875854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.18176000118255614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.17889280319213868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.1934272050857544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.19833600521087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.20532479286193847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.23401598930358886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.2515775918960571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.304531192779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.32917759418487547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.39598081111907957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.4830463886260986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.6412032127380372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.7972864151000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.1028800010681152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.4626239776611327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,balanced,0.10242666800816853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,balanced,0.11489066481590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,balanced,0.11583999792734782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,balanced,0.11528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,balanced,0.11505066355069478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,balanced,0.11448533336321513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,balanced,0.11834667126337688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,balanced,0.11638399958610535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,balanced,0.11684266726175944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,balanced,0.11818666259447734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,balanced,0.1242026686668396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,balanced,0.12423466642697652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,balanced,0.12718400359153748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,balanced,0.13795733451843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,balanced,0.1344480017820994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,balanced,0.16513066490491232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,balanced,0.15334933002789816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,balanced,0.21517332394917807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,balanced,0.20143999656041464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,balanced,0.3408799966176351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,balanced,0.29792000850041706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,balanced,0.43831467628479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,balanced,0.4894560178120931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,balanced,0.03323733309904734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,power_law_1.01,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,balanced,0.02942933390537898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,balanced,0.029743999242782593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,power_law_1.01,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,balanced,0.029365333418051403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,balanced,0.029530666768550873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,balanced,0.031194667021433514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,balanced,0.02938666691382726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,balanced,0.031045332551002502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,balanced,0.03120533376932144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,balanced,0.03196266790231069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,balanced,0.03931200007597605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,balanced,0.061349332332611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,balanced,0.0846560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,balanced,0.09685333569844563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,balanced,0.1311360001564026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,balanced,0.1370560030142466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,power_law_1.01,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,power_law_1.01,0.08439040184020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,power_law_1.01,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,power_law_1.01,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,power_law_1.01,0.08767359852790832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,power_law_1.01,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,power_law_1.01,0.09247999787330627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,power_law_1.01,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,power_law_1.01,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,power_law_1.01,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,power_law_1.01,0.13165440559387206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,power_law_1.01,0.15403519868850707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,power_law_1.01,0.1907520055770874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,power_law_1.01,0.21852800846099854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,power_law_1.01,0.2863935947418213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,power_law_1.01,0.35341439247131345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,power_law_1.01,0.49013118743896483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,power_law_1.01,0.6226880073547363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,power_law_1.01,0.8801407814025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,power_law_1.01,1.1362751960754394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,balanced,0.03801066676775614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,balanced,0.07806933422883351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,balanced,0.11867733796437581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,balanced,0.20230400562286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,balanced,0.20278932650883993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,balanced,0.2067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,balanced,0.2067413330078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,balanced,0.20675732692082724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,balanced,0.2076746622721354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,balanced,0.21034133434295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,balanced,0.21252799034118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,balanced,0.21411200364430746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,balanced,0.2167146603266398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,balanced,0.2230773369471232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,balanced,0.22900799910227457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,balanced,0.23314666748046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,balanced,0.2515786687533061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,balanced,0.2702239950497945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,balanced,0.3071253299713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,balanced,0.3245760003725688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,balanced,0.40170665582021076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,balanced,0.46242133776346844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,balanced,0.6574293375015259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,balanced,0.7433706919352213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,balanced,1.0498186747233074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,balanced,1.3703200022379558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,power_law_1.2,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,power_law_1.2,0.10776319503784179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,power_law_1.2,0.17303680181503295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,power_law_1.2,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,power_law_1.2,0.18750079870223998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,power_law_1.2,0.1949952006340027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,power_law_1.2,0.1970047950744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,power_law_1.2,0.1999295949935913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,power_law_1.2,0.2073215961456299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,power_law_1.2,0.20915839672088624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,power_law_1.2,0.21395199298858641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,power_law_1.2,0.22647039890289306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,power_law_1.2,0.23423359394073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,power_law_1.2,0.23343360424041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,power_law_1.2,0.2576256036758423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.2753535985946655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.3230144023895264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.34805119037628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.43240962028503416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.5292543888092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,power_law_1.2,0.6849728107452393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,balanced,0.0646666685740153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,balanced,0.08902933200200398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,balanced,0.10965866843859355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,balanced,0.10977066556612651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,balanced,0.11354133486747742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,balanced,0.11143466830253601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,balanced,0.1137600044409434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,balanced,0.11318932970364888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,balanced,0.1158026655515035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,balanced,0.11783466736475627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,balanced,0.12075733145078023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,balanced,0.12363200386365254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,balanced,0.12739200393358865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,balanced,0.13404267032941183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,balanced,0.14257066448529562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,balanced,0.1638826628526052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,balanced,0.178874671459198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,balanced,0.21124800046284994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,balanced,0.24519999821980795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,balanced,0.35843201478322345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,balanced,0.44139734903971356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,balanced,0.62937064965566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,power_law_1.2,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,balanced,0.8126346270243326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,power_law_1.2,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,power_law_1.2,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,power_law_1.2,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,power_law_1.2,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,power_law_1.2,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,power_law_1.2,0.10856319665908813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,power_law_1.2,0.11139199733734131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,power_law_1.2,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,power_law_1.2,0.116428804397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,power_law_1.2,0.11274240016937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,power_law_1.2,0.12174719572067261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,power_law_1.2,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,power_law_1.2,0.12179839611053467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,power_law_1.2,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,power_law_1.2,0.13990399837493897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,power_law_1.2,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,power_law_1.2,0.16049280166625976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,power_law_1.2,0.8247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.2,0.17108479738235474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.2,0.19043200016021727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.2,0.22815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.2,0.3051455974578857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.2,0.38407039642333984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.2,0.5446464061737061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.2,0.7834368228912354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.2,1.0064703941345214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.2,1.466700839996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,power_law_1.2,1.1304256439208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,power_law_1.2,1.4312447547912597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.0673471987247467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.07110400199890136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.09624320268630981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.1610368013381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.18353919982910155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.22748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.2578367948532104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.3514303922653198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.4830783843994141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.658790397644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.8488960266113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,power_law_1.2,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,power_law_1.2,0.04840959906578064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,power_law_1.2,0.04953599870204926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,power_law_1.2,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,power_law_1.2,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,power_law_1.2,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.2,0.09959679841995239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.2,0.12029440402984619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.2,0.14324480295181274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.2,0.2043328046798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.2,0.2435904026031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.2,0.3710783958435059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.2,0.5488831996917725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.2,0.701196813583374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.2,1.1564288139343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,power_law_1.2,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,power_law_1.2,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,power_law_1.2,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,power_law_1.2,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,power_law_1.2,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.2,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.2,0.15542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.2,0.2107072114944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.2,0.25250558853149413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.2,0.42639999389648436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,balanced,0.02535466601451238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,balanced,0.025455998877684276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,balanced,0.025066666305065155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,balanced,0.02382933348417282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,balanced,0.025061334172884624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,balanced,0.02714666724205017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,balanced,0.027130665878454845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,balanced,0.02565866708755493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,balanced,0.025439999997615814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,balanced,0.027061333258946735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,balanced,0.029493334392706554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,balanced,0.03385066737731298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,balanced,0.03457599878311157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,balanced,0.039408000806967415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,balanced,0.08550399541854858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,balanced,0.09665066997210185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.2,0.578003215789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.2,0.794758415222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.2,1.0618240356445312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,balanced,0.034416000048319496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,balanced,0.03665599972009659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,balanced,0.037978666524092354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,balanced,0.05075199902057648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,balanced,0.05132266879081726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,balanced,0.06061866879463196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,balanced,0.07981333136558533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,balanced,0.0979306697845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,balanced,0.11395733555157979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,balanced,0.17324266831080118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,balanced,0.23002666234970093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,balanced,0.27427200476328534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,balanced,0.04632000128428141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,balanced,0.07842666904131572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,balanced,0.10936533411343892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,balanced,0.11202133695284526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,balanced,0.11388799548149109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,balanced,0.11520000298817952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,balanced,0.11497599879900615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,balanced,0.11693867047627766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,balanced,0.12133866548538208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,balanced,0.12154133121172588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,balanced,0.12459199627240498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,balanced,0.13471999764442444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,balanced,0.13714133699735007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,balanced,0.14729600151379904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,balanced,0.16250133514404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,balanced,0.17537599802017212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,balanced,0.20336000124613443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,balanced,0.2330933411916097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,balanced,0.30513066053390503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,balanced,0.3650826613108317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,balanced,0.5119733413060507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,balanced,0.6270826657613119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,balanced,0.8970826466878256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,balanced,1.1489386558532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.11698559522628785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.13749120235443116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.18229119777679442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.19157760143280028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.2561216115951538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.2689215898513794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.23688321113586425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.2629184007644653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.24908800125122071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.26888959407806395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.2711935997009277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.28520960807800294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.28499200344085696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.29690239429473875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.3099456071853638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.3566848039627075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.39643518924713134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.4520383834838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.5444479942321777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.6277503967285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.8025728225708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.025376033782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.3343999862670899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.6404224395751954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.07343999743461609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.13017599582672118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.1583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.2292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.2409600019454956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.25128960609436035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.2517568111419678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.26019198894500734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2665855884552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.2726016044616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.27557759284973143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.281004810333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.2990528106689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.3117248058319092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.33045759201049807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.37781760692596433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.41020798683166504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.4891456127166748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.5684480190277099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.7375552177429199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.8845824241638184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.188697624206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,1.5014847755432128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.0918720245361326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,2.691257667541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,balanced,0.07693333427111308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,balanced,0.08514666557312012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,balanced,0.09967999656995137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,balanced,0.1246453324953715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,balanced,0.12918933232625326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,balanced,0.13159466783205667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,balanced,0.12985066572825113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,balanced,0.1353493332862854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,balanced,0.13691733280817667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,balanced,0.139573335647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,balanced,0.1425333321094513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,balanced,0.1436906655629476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,balanced,0.14979733030001322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,balanced,0.14590932925542197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,balanced,0.14784533778826395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,balanced,0.15226133664449057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,balanced,0.16569599509239197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,balanced,0.18038400014241537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,balanced,0.20799465974171957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,balanced,0.2418880065282186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,balanced,0.3439360062281291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,balanced,0.4109813372294108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,balanced,0.582255999247233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,balanced,0.7554773489634196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,balanced,1.0886507034301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,balanced,1.433685302734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,balanced,0.03719999889532725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.052890668312708534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.07321066657702129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.08928533395131429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.1127359966437022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.13522666692733765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.18422933419545492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.22276800870895386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,balanced,0.08735466996828715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,balanced,0.12562132875124613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,balanced,0.1923840045928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,balanced,0.33375998338063556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,balanced,0.5986080169677734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,balanced,0.8426240285237631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,balanced,0.8478933175404867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,balanced,0.8496267000834147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,balanced,0.851919968922933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,balanced,0.8565866947174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,balanced,0.8642346858978271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,balanced,0.8720800081888834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,balanced,0.8752480347951254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,balanced,0.881114641825358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,balanced,0.8911146322886149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,balanced,0.8982453346252441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,balanced,0.9130880037943522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,balanced,1.120789368947347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,balanced,0.9786293506622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,balanced,1.2397546768188477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,balanced,1.1203093528747559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,balanced,1.79367462793986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,balanced,1.372752030690511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,balanced,2.103402614593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,balanced,2.075984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,balanced,3.1442346572875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,balanced,3.4468161265055337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.09930239915847779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.1406399965286255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.14238719940185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.11461119651794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.12391680479049683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.12802560329437257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.1318400025367737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.13910399675369262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.14115840196609497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.14897919893264772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.15000319480895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.17365119457244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.18414080142974854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.1642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.1857856035232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.18805760145187378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.22900478839874266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.23675520420074464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.30217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.3430912017822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,power_law_1.2,0.5039807796478272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,power_law_1.2,0.5613056182861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,power_law_1.2,0.7512703895568847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,power_law_1.2,0.9180671691894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,balanced,0.05693866809209188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,balanced,0.05774933099746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,balanced,0.059279998143514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,balanced,0.0621919979651769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,balanced,0.06677333513895671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,balanced,0.06896533568700154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,balanced,0.09307733178138733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,balanced,0.12384532888730367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,balanced,0.13868266344070435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,balanced,0.1860640048980713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,balanced,0.21776533126831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,balanced,0.313210666179657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,balanced,0.37522133191426593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,balanced,0.5333919922510783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,balanced,0.6831786632537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.09842560291290284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.12632319927215577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.1404863953590393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.1368384003639221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.13752319812774658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.14805760383605956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.16247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.18071039915084838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.18821760416030883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.19761279821395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.2524224042892456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.2979775905609131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,power_law_1.01,0.39507839679718015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,power_law_1.01,0.47431039810180664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,power_law_1.01,0.5977280139923096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,power_law_1.01,0.7475264072418213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,power_law_1.01,1.0584832191467286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,power_law_1.01,1.390233612060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,power_law_1.01,1.9965887069702148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,power_law_1.2,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.2,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.2,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.2,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.2,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.2,0.16046080589294434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.2,0.21790080070495604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.2,0.3032576084136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,power_law_1.01,2.743155288696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.2,0.4258431911468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,balanced,0.05229333539803823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,balanced,0.05332266787687937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,balanced,0.08449600140253703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,balanced,0.11593600114186604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,balanced,0.1195199986298879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,balanced,0.12181867162386577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,balanced,0.11941867073376973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,balanced,0.12131733695665996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,balanced,0.12220266461372375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,balanced,0.12150933345158894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,balanced,0.12380266189575195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,balanced,0.12405866384506226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,balanced,0.1365333298842112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,balanced,0.13886933525403342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,balanced,0.14409599701563516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,balanced,0.1583199997742971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,balanced,0.16523200273513794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,balanced,0.17887999614079794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,balanced,0.19357866048812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,balanced,0.22828267018000284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,balanced,0.2541866699854533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,balanced,0.33766400814056396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,balanced,0.4012586673100789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,balanced,0.5623946587244669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,balanced,0.7275733153025309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.13964799642562867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.1919167995452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.3089215993881226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.443225622177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5765632152557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.7884543895721435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8300352096557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8559935569763184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.8763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.8941311836242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9223872184753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9369728088378906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.944979190826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.993619155883789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0094464302062989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.0822976112365723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.151308822631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,balanced,0.053557331363360085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.3106752395629884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,balanced,0.07685866455237071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,balanced,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,balanced,0.0759200006723404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,balanced,0.07600000003973643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,balanced,0.07938666641712189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.3547200202941894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,1.6259775161743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,1.8773056030273438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,balanced,0.0841493308544159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,balanced,0.08647466699282329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,balanced,0.08675733208656311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,balanced,0.09206933776537578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,balanced,0.09410666426022847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,balanced,0.10563733180363973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,balanced,0.11596266428629558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,2.4367488861083983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,balanced,0.1328000028928121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,balanced,0.14920533696810404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,balanced,0.18930667638778687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,balanced,0.22781866788864136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,balanced,0.2973173260688782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,balanced,0.36324799060821533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,2.9114879608154296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,3.9164161682128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,4.93048324584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,power_law_1.2,0.08047999739646912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,power_law_1.2,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,power_law_1.2,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,power_law_1.2,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,power_law_1.2,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,power_law_1.2,0.08196480274200439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,power_law_1.2,0.08385279774665833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,power_law_1.2,0.08638079762458802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,power_law_1.2,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,power_law_1.2,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,power_law_1.2,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,power_law_1.2,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,power_law_1.2,0.11578880548477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,power_law_1.2,0.14748799800872803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,power_law_1.2,0.14342399835586547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,power_law_1.2,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,power_law_1.2,0.20200960636138915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,power_law_1.2,0.28187520503997804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,power_law_1.2,0.3444672107696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,power_law_1.2,0.39886720180511476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,power_law_1.2,0.603872013092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,power_law_1.2,0.8078463554382325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,balanced,0.039274667700131737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,balanced,0.04435733457406362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,balanced,0.04541333516438802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,balanced,0.04957866668701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,balanced,0.05154666801293691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,power_law_1.2,1.1101119995117188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,balanced,0.06899733344713847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,balanced,0.08683733145395915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,balanced,0.10601600011189778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,balanced,0.12040533622105916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,balanced,0.1605226695537567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,balanced,0.19171200195948282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,balanced,0.27089067300160724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,balanced,0.3509226640065511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,power_law_1.2,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,power_law_1.2,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,power_law_1.2,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,power_law_1.2,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,power_law_1.2,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,power_law_1.2,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,power_law_1.2,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,power_law_1.2,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,power_law_1.2,0.11908479928970336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,power_law_1.2,0.17493120431900025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,power_law_1.2,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,power_law_1.2,0.26023681163787843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,power_law_1.2,0.40320639610290526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,power_law_1.2,0.4681727886199951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,balanced,0.10225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,balanced,0.15189866224924722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,balanced,0.17802133162816366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,balanced,0.1804479956626892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,balanced,0.18336000045140585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,balanced,0.18193066120147705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,balanced,0.1829599936803182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,balanced,0.18313600619633993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,balanced,0.18471999963124594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,balanced,0.18690667549769083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,balanced,0.18508267402648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,balanced,0.19054933389027914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,balanced,0.19318934281667074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,balanced,0.1970133384068807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,balanced,0.21532267332077026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,balanced,0.2113493283589681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,balanced,0.2431946595509847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,balanced,0.23916266361872354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,balanced,0.311082661151886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,balanced,0.3008693257967631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,balanced,0.4236373504002889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,balanced,0.45504534244537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,balanced,0.6863093376159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,balanced,0.7629280090332031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,power_law_1.2,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,power_law_1.2,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,power_law_1.2,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,power_law_1.2,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.2,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.2,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.2,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.2,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,balanced,0.059861332178115845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,balanced,0.08290133376916249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,balanced,0.09966400265693665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,balanced,0.10087466239929199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,balanced,0.10007466872533162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,balanced,0.10236266255378723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,balanced,0.1049013336499532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,balanced,0.10569600264231364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,balanced,0.11384000380833943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,balanced,0.11334932843844096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,balanced,0.11706133683522542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,balanced,0.12546666463216147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.13083733121554056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.1466506620248159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.16214932998021445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.19035732746124268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.22121065855026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.27777600288391113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.33904532591501874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.447653333346049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.5806613365809122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.2,0.11297919750213622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.2,0.17089920043945311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.2,0.20336639881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.2,0.26632959842681886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.2,0.393503999710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,power_law_1.01,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,power_law_1.01,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,power_law_1.01,0.08872960209846496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,power_law_1.01,0.0876416027545929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,power_law_1.01,0.08234239816665649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,power_law_1.01,0.08776959776878357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,power_law_1.01,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,power_law_1.01,0.08847360014915466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,power_law_1.01,0.08899199962615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,power_law_1.01,0.08847360014915466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,power_law_1.01,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,power_law_1.01,0.09230080246925354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,power_law_1.01,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,power_law_1.01,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,power_law_1.01,0.10843520164489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,power_law_1.01,0.12241920232772827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.01,0.13019520044326782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.01,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,balanced,0.090229332447052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,balanced,0.11157866319020589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,balanced,0.15989333391189575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,balanced,0.2072640061378479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,balanced,0.19808000326156616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,balanced,0.19960000117619833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,balanced,0.20080532630284628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,balanced,0.2020853360493978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,balanced,0.20522665977478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,balanced,0.20803733666737875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,balanced,0.21067200104395548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,balanced,0.21985065937042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,balanced,0.2225760022799174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,balanced,0.23193599780400595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,balanced,0.2537440061569214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,balanced,0.2741760015487671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,balanced,0.32089600960413617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.01,0.1749119997024536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,balanced,0.3559199968973796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,balanced,0.45190401871999103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,balanced,0.5367946624755859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.01,0.22685439586639405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,balanced,0.7280373573303223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,balanced,0.9927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.01,0.27638399600982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,balanced,1.3676212628682454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.01,0.38913280963897706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,balanced,1.8229014078776042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.01,0.5311168193817138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.01,0.7484416007995606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.01,1.0122688293457032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,power_law_1.01,0.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,power_law_1.01,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,power_law_1.01,0.0225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,power_law_1.01,0.022572800517082214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,power_law_1.01,0.02383359968662262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,power_law_1.01,0.02356480062007904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,power_law_1.01,0.023923200368881226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,power_law_1.01,0.02391040027141571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,power_law_1.01,0.027225598692893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,power_law_1.01,0.02805759906768799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,power_law_1.01,0.02998400032520294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,power_law_1.01,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.07824640274047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.09262719750404358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.13646080493927001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.16573439836502074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,balanced,0.07075199981530507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,balanced,0.07278400162855785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,balanced,0.07449066638946533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,balanced,0.10643200079600017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,balanced,0.13012267152468363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,balanced,0.12993599971135458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,balanced,0.13408000270525613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,balanced,0.13125333189964294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,balanced,0.1337440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,balanced,0.1341866652170817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,balanced,0.13573333621025085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,balanced,0.13919466733932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,balanced,0.1418506701787313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,balanced,0.14839466412862143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,balanced,0.15036267042160034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,balanced,0.15447466572125754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,balanced,0.1667520006497701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,balanced,0.1798400084177653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,balanced,0.2111253341039022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,balanced,0.24037333329518637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,balanced,0.31031467517217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,balanced,0.37595733006795246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,balanced,0.5436213413874308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,balanced,0.6740266482035319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,balanced,0.9781493345896403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,balanced,1.2755680084228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.10387200117111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.1672127962112427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.17982079982757568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.18654719591140748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.19055360555648804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.19201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.20135040283203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.2101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.20984959602355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.2149951934814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.22942719459533692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.23530879020690917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.24635519981384277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.27968640327453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.27708160877227783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.3167167901992798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.3415872097015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.4063231945037842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.2,0.46898560523986815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.2,0.5753600120544433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.2,0.6801536083221436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.2,0.9191103935241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.2,1.156454372406006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,power_law_1.01,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,power_law_1.01,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,power_law_1.01,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,power_law_1.01,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,power_law_1.01,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,power_law_1.01,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,power_law_1.01,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,power_law_1.01,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,power_law_1.01,0.125382399559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,balanced,0.037263999382654824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,power_law_1.01,0.13932160139083863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,balanced,0.036917333801587425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,power_law_1.01,0.16232320070266723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,balanced,0.06583466629187266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,balanced,0.07855466504891713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,balanced,0.09411733349164327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,balanced,0.12181333700815837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,balanced,0.14495467146237692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,balanced,0.20172800620396933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,balanced,0.2704746723175049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,balanced,0.403989315032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,balanced,0.5217546621958414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,power_law_1.01,0.1898687958717346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,power_law_1.01,0.2760512113571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,balanced,0.054341331124305725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,balanced,0.0687360018491745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,balanced,0.08180266618728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,balanced,0.08356799681981404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,balanced,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,balanced,0.08699199557304382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,balanced,0.09074667096138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,balanced,0.09085333347320557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,balanced,0.09036800265312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,balanced,0.0928106705347697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,balanced,0.09345066547393799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,balanced,0.10099732875823975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,balanced,0.10297600428263347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,balanced,0.10582933823267619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,balanced,0.1113813320795695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,balanced,0.11672533551851909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,balanced,0.12995200355847678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,balanced,0.14103466272354126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,balanced,0.1772586703300476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,balanced,0.1980853279431661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,balanced,0.24843732515970865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,balanced,0.29402132829030353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,balanced,0.4077599843343099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,balanced,0.5086719989776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,power_law_1.01,0.3197824001312256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,power_law_1.01,0.4379583835601807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,power_law_1.01,0.5955647945404052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,power_law_1.01,0.8708031654357911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,power_law_1.01,1.2006976127624511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,balanced,0.03677333394686381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,balanced,0.037087999284267426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,balanced,0.050197333097457886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,balanced,0.048623998959859215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,balanced,0.05032533407211304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,balanced,0.056474665800730385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,balanced,0.05840000013510386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,balanced,0.06262933214505513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,balanced,0.0666240006685257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.08273600041866302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.0915786623954773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.12241599957148235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.1416906714439392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.18106667200724283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.21433599789937338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.29763199885686237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.37542935212453205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,balanced,0.051541333397229515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,balanced,0.045466666420300804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,balanced,0.047466665506362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,balanced,0.047584002216657005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,balanced,0.05144000053405762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,balanced,0.07565866907437642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,balanced,0.08873066306114197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,balanced,0.11770133177439372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,balanced,0.13615467151006064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,balanced,0.18120000759760538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,balanced,0.22028799851735434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,balanced,0.2976213296254476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,balanced,0.3758933146794637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,balanced,0.08057066798210144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,balanced,0.0824480007092158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,balanced,0.08859733740488689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,balanced,0.09949333469072978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,balanced,0.1325653294722239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,balanced,0.18673600753148398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,balanced,0.18810667594273886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,balanced,0.18966933091481528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,balanced,0.18915200233459473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,balanced,0.19182932376861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,balanced,0.1964319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,balanced,0.2032159964243571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,balanced,0.20230400562286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,balanced,0.20508799950281778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,balanced,0.21289600928624472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,balanced,0.22066134214401245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,balanced,0.22965866327285767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,balanced,0.25845867395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,balanced,0.2869066596031189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,balanced,0.3421333233515422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,balanced,0.3949813445409139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,balanced,0.512234648068746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,balanced,0.6244959831237793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,balanced,0.9206346670786539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,balanced,1.145466645558675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,balanced,1.6876479784647624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,balanced,2.2044533093770347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,power_law_1.2,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,power_law_1.2,0.10424959659576416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,power_law_1.2,0.1312064051628113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,power_law_1.2,0.16252800226211547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,power_law_1.2,0.23208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,power_law_1.2,0.2393728017807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,power_law_1.2,0.24828801155090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,power_law_1.2,0.2498687982559204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,power_law_1.2,0.2659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,power_law_1.2,0.27285759449005126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,power_law_1.2,0.27802879810333253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,power_law_1.2,0.2932159900665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,power_law_1.2,0.3046015977859497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,power_law_1.2,0.3117120027542114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,power_law_1.2,0.32773120403289796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,power_law_1.2,0.36161279678344727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,power_law_1.2,0.38098559379577634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,power_law_1.2,0.4388864040374756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,power_law_1.2,0.4869376182556152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,power_law_1.2,0.6023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,power_law_1.2,0.7057472229003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,power_law_1.2,0.895628833770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,power_law_1.2,1.0972288131713868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,power_law_1.2,1.500979232788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,power_law_1.2,1.9001087188720702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.2,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.2,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.2,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.2,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.2,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.2,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.2,0.09964799880981445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.2,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.2,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.2,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.2,0.24678399562835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08443520069122315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.08565120100975036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.09971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.12453759908676147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.13747199773788452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.15480320453643798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.19771519899368287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.2159679889678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.28933119773864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.36688640117645266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,power_law_1.01,0.5058047771453857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,power_law_1.01,0.6223167896270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.09702399969100953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.09222400188446045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.09271039962768554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.09485440254211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.10369919538497925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.10820480585098266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.1154047966003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.13277440071105956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.1378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.17844480276107788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.2258944034576416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.2860543966293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,power_law_1.2,0.3246272087097168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,power_law_1.2,0.4073023796081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,power_law_1.2,0.6329408168792725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,power_law_1.2,0.8897279739379883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,power_law_1.2,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,power_law_1.2,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,power_law_1.2,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,power_law_1.2,0.06254079937934875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,power_law_1.2,0.06507520079612732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,power_law_1.2,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,power_law_1.2,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,power_law_1.2,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,power_law_1.2,0.06832000017166137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,power_law_1.2,0.07035520076751708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,power_law_1.2,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,power_law_1.2,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,power_law_1.2,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,power_law_1.2,0.10796799659729003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,power_law_1.2,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,power_law_1.2,0.1464192032814026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,power_law_1.2,0.17935359477996826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,power_law_1.2,0.2351423978805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,power_law_1.2,0.2996223926544189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,power_law_1.2,0.41216001510620115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,power_law_1.2,0.6245440006256103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,power_law_1.2,0.664735984802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,power_law_1.2,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,power_law_1.2,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,power_law_1.2,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,power_law_1.2,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,power_law_1.2,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,power_law_1.2,0.06862720251083373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,power_law_1.2,0.07996799945831298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,power_law_1.2,0.09456639885902404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,power_law_1.2,0.10503679513931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,power_law_1.2,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,power_law_1.2,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,power_law_1.2,0.16202239990234374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,power_law_1.2,0.19597439765930175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,power_law_1.2,0.2549184083938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,power_law_1.2,0.33483519554138186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,power_law_1.2,0.45224318504333494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,power_law_1.2,0.5990208148956299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,power_law_1.2,0.888326358795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,power_law_1.2,1.0955072402954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.050291198492050174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.0782975971698761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.10992640256881714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.19064320325851442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.22811520099639893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.33165440559387205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,power_law_1.01,0.41455998420715334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,power_law_1.01,0.5368512153625489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,power_law_1.01,0.6328000068664551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,balanced,0.04004266609748205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,balanced,0.06176533301671346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,balanced,0.08071466783682506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,balanced,0.08203200002511342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,balanced,0.08343999584515889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,balanced,0.08307200173536937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,balanced,0.08383466800053914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,balanced,0.0846453309059143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,balanced,0.0844106674194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,balanced,0.08743466933568318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,balanced,0.08820799986521403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,balanced,0.09686400492986043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,balanced,0.10007466872533162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,balanced,0.10825066765149434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,balanced,0.11521599690119426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,balanced,0.13013333082199097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,balanced,0.1441386640071869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,balanced,0.17630932728449503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,balanced,0.20907733837763467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,balanced,0.2765973409016927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,balanced,0.3301333387692769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,balanced,0.46938133239746094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,balanced,0.5874613523483276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.13574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.20102400779724122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.32099840641021726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.5278143882751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.01,0.7259391784667969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.2044095993041992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.01,1.3262016296386718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.01,1.3485952377319337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.01,1.3889984130859374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.01,1.4110015869140624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.01,1.4534655570983888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.01,1.4890751838684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.01,1.506169605255127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.01,1.5654208183288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.01,1.6060096740722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.01,1.649817657470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.01,1.7340095520019532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.01,1.8755456924438476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.01,1.9147008895874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.01,2.174611282348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.01,2.3919296264648438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.01,2.9173120498657226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.01,2.942905616760254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.01,3.565804672241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.01,4.633062362670898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.01,6.380902481079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.01,7.315679931640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,balanced,0.08591999610265096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,balanced,0.09325333436330159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,balanced,0.09658132990201314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,balanced,0.09554133812586467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,balanced,0.09554133812586467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,balanced,0.09646933277448018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,balanced,0.0972160001595815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,balanced,0.09731200337409973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,balanced,0.10071466366449992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,balanced,0.10602133472760518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,balanced,0.10603200395901997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,balanced,0.10674132903416951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,balanced,0.11570133765538533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.11953600247701009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.1439359982808431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.14959466457366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.1824959913889567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.19819732507069907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.2735893328984578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.31194667021433514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.425872008005778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.5240480105082194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,power_law_1.2,0.0555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,power_law_1.2,0.06168320178985596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,power_law_1.2,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,power_law_1.2,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,power_law_1.2,0.07989760041236878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,power_law_1.2,0.08365439772605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,power_law_1.2,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,power_law_1.2,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,power_law_1.2,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,power_law_1.2,0.08473600149154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,power_law_1.2,0.08746880292892456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,power_law_1.2,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,power_law_1.2,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,power_law_1.2,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,power_law_1.2,0.1288256049156189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.2,0.14023040533065795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.2,0.1740928053855896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.2,0.1859392046928406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.2,0.2395008087158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.2,0.3038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.2,0.42223358154296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.2,0.5314559936523438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.2,0.756006383895874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.2,0.943295955657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,balanced,0.04292800029118856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.08860799670219421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.10878933469454448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.12904533743858337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.17485332489013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.21534399191538492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.30726399024327594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.39771199226379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,balanced,0.05086400111516317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,balanced,0.07037333150704701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,balanced,0.06816000243028005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,balanced,0.07011733452479045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,balanced,0.07240533332029979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,balanced,0.07236266632874806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,balanced,0.08368000388145447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,balanced,0.08249600231647491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,balanced,0.08710400263468425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,balanced,0.09860799709955852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.10745066404342651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.1443946659564972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.1764799952507019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.20770132541656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.2695680061976115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.34306132793426514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.4710079828898112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.6053280035654703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,power_law_1.01,0.047686401009559634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,power_law_1.01,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,power_law_1.01,0.06216319799423218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,power_law_1.01,0.06320639848709106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,power_law_1.01,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,power_law_1.01,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,power_law_1.01,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,power_law_1.01,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,power_law_1.01,0.09382399916648865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,power_law_1.01,0.10988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,power_law_1.01,0.13307520151138305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,power_law_1.01,0.16750719547271728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,power_law_1.01,0.21527678966522218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,power_law_1.01,0.2666944026947021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,power_law_1.01,0.3711616039276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,power_law_1.01,0.44953598976135256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,power_law_1.01,0.6801280021667481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,power_law_1.01,0.8424192428588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.0932864010334015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.10496000051498414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.13166719675064087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.16453759670257567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.23096320629119874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.2904576063156128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.33843839168548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.4131455898284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.4242368221282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.43751678466796873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.44965758323669436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.4664447784423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.47450881004333495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.48718719482421874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.5011712074279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.5217472076416015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.5361536026000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.5916224002838135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.636729621887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,0.7357696056365967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,0.8227840423583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,1.0148863792419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,1.2003199577331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,1.5825152397155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.01,1.9561344146728517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.01,2.707904052734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.01,3.5729598999023438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.2,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,power_law_1.2,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.2,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,power_law_1.2,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.2,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.2,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.2,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.2,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,power_law_1.2,0.0773248016834259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.2,0.1259328007698059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,power_law_1.2,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.2,0.1486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.2,0.19372160434722902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.2,0.20789759159088134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.2,0.27302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.2,0.33914239406585694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.2,0.13035520315170288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.2,0.4825471878051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.2,0.1798848032951355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.2,0.19980159997940064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.2,0.6159808158874511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.2,0.3054784059524536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.2,0.38657920360565184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.2,0.6425663948059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.2,0.8591103553771973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.2,1.175379180908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.2,1.4263936042785645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,balanced,0.032069332897663116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,balanced,0.038586666186650596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,balanced,0.060517330964406334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,balanced,0.05979733169078827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,balanced,0.061306665341059365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,balanced,0.07158400118350983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,balanced,0.07061866422494252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,balanced,0.07287466526031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,balanced,0.07979733248551686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,balanced,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,balanced,0.10220799843470256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,balanced,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,balanced,0.1515733301639557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,balanced,0.16870933771133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,balanced,0.21779199441274008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,balanced,0.27596267064412433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,balanced,0.3835306564966838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,balanced,0.495471994082133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,power_law_1.2,0.05820159912109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,power_law_1.2,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,power_law_1.2,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,power_law_1.2,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,power_law_1.2,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,power_law_1.2,0.1158784031867981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,power_law_1.2,0.1501952052116394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,power_law_1.2,0.15145599842071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,power_law_1.2,0.1552896022796631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,power_law_1.2,0.17114880084991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,power_law_1.2,0.15500160455703735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,power_law_1.2,0.16138880252838134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,power_law_1.2,0.15619839429855348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,power_law_1.2,0.16010240316390992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,power_law_1.2,0.16860159635543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,power_law_1.2,0.17537280321121215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,power_law_1.2,0.1861567974090576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,power_law_1.2,0.21137280464172364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,power_law_1.2,0.2151616096496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,power_law_1.2,0.24657280445098878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,power_law_1.2,0.2735487937927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,power_law_1.2,0.35433599948883054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,power_law_1.2,0.3551552057266235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,power_law_1.2,0.4557759761810303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,power_law_1.2,0.564902400970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,power_law_1.2,0.7525824069976806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,power_law_1.2,1.0540224075317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,balanced,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,balanced,0.04001600046952566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,balanced,0.04139200101296107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,balanced,0.054773335655530296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,balanced,0.07633600135644276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,balanced,0.09341333309809367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,balanced,0.10748266180356343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,balanced,0.14019733667373657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,balanced,0.16755733887354532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,balanced,0.22807466983795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,balanced,0.286463995774587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,balanced,0.40213334560394287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,balanced,0.5038613478342692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,balanced,0.09753599762916565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,balanced,0.14247467120488486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,balanced,0.21422932545344034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,balanced,0.288975993792216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,balanced,0.28756266832351685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,balanced,0.2938613295555115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,balanced,0.2940160036087036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,balanced,0.29851200183232623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,balanced,0.3068480094273885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,balanced,0.3102773427963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,balanced,0.3183893362681071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,balanced,0.3249866763750712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,balanced,0.33136532704035443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,balanced,0.3529119888941447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,balanced,0.4002133210500081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,balanced,0.4362666606903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,balanced,0.5101973215738932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,balanced,0.5732479890187582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,balanced,0.7323520183563232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,balanced,0.9277013142903646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,balanced,1.2194560368855794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,balanced,1.6387732823689778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,balanced,2.33405335744222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,balanced,3.0329386393229165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,balanced,0.06467199822266896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,balanced,0.09009599685668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,balanced,0.18472532431284586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,balanced,0.2480319937070211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,balanced,0.24992533524831137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,balanced,0.25117333730061847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,balanced,0.24918399254480997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,balanced,0.2518506646156311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,balanced,0.2560960054397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,balanced,0.2586933374404907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,balanced,0.2627413272857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,balanced,0.271178662776947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,balanced,0.27242134014765423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,balanced,0.285258670647939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,balanced,0.29791466395060223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.3158666690190633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.3487093448638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.38254400094350177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.45123199621836346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.5014933347702026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.6201066573460897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,balanced,0.8025279839833578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,balanced,1.0193760395050049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,balanced,1.3647252718607585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,balanced,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,balanced,0.14802666505177817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,balanced,0.2479520042737325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,balanced,0.4429440100987752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,balanced,0.8302079836527506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,balanced,1.6179626782735188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,balanced,1.6228747367858887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,balanced,1.615664005279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,balanced,1.6220426559448242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,balanced,1.62664000193278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,balanced,1.6471467018127441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,balanced,1.6597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,balanced,1.662698745727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,balanced,1.666122595469157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,balanced,1.6818240483601887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,balanced,1.6917813618977864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,balanced,1.7140266100565593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,balanced,1.7722773551940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,balanced,1.8137332598368328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,balanced,1.896959940592448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,balanced,2.0090506871541343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,balanced,2.178581396738688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,balanced,2.321194648742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,balanced,2.9783252080281577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,balanced,3.0298026402791343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,balanced,4.658405303955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,balanced,4.989935874938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,power_law_1.2,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,power_law_1.2,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,power_law_1.2,0.09492480158805847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,power_law_1.2,0.11703039407730102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,power_law_1.2,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,power_law_1.2,0.18388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,power_law_1.2,0.19301120042800904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,power_law_1.2,0.19613440036773683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,power_law_1.2,0.20531840324401857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,power_law_1.2,0.21122560501098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,power_law_1.2,0.21539840698242188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,power_law_1.2,0.2207808017730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,power_law_1.2,0.23173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,power_law_1.2,0.2453376054763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,power_law_1.2,0.2554944038391113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,power_law_1.2,0.27762560844421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,power_law_1.2,0.33035519123077395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,power_law_1.2,0.36537599563598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,power_law_1.2,0.45255680084228517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,power_law_1.2,0.5321536064147949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,power_law_1.2,0.6956223964691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,power_law_1.2,0.9217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,power_law_1.2,1.2316287994384765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,power_law_1.2,1.5752384185791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,power_law_1.2,2.175712013244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,power_law_1.2,2.786956787109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,balanced,0.05423999826113383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,balanced,0.07529066503047943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,balanced,0.09999466935793559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,balanced,0.14961066842079163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,balanced,0.2469600041707357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,balanced,0.4448106686274211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,balanced,0.642735997835795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,balanced,0.6402080059051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,balanced,0.6437919934590658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,balanced,0.6465813318888346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,balanced,0.6489599943161011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,balanced,0.6571840047836304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,balanced,0.6636159817377726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,balanced,0.6661813259124756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,balanced,0.6760000387827555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,balanced,0.6844800313313802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,balanced,0.7018400033315023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,balanced,0.732314666112264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,balanced,0.7527999877929688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,balanced,0.8084853490193685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,balanced,0.8508640130360922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,balanced,0.9648799896240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,balanced,1.0583893458048503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,balanced,1.26637864112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,balanced,1.548074722290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,balanced,2.00382932027181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,balanced,2.5589653650919595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,balanced,0.04506133496761322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,balanced,0.07361066838105519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,balanced,0.07668266693751018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,balanced,0.0766133318344752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,balanced,0.0764213353395462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,balanced,0.07696533203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,balanced,0.08006933331489563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,balanced,0.08228800197442372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,balanced,0.09239466985066731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,balanced,0.09501866499582927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,balanced,0.10376532872517903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,balanced,0.11652800440788269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,balanced,0.13197867075602213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,balanced,0.15528000394503275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,balanced,0.17643199364344278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,balanced,0.25014932950337726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,balanced,0.3043573300043742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,balanced,0.43459200859069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,balanced,0.5651520093282064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,balanced,0.8273013432820638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,balanced,1.076474666595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,balanced,0.14915200074513754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,balanced,0.24547199408213297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,balanced,0.43748267491658527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,balanced,0.8263093630472819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,balanced,1.6079039573669434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,balanced,3.1410719553629556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,balanced,4.686938603719075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,balanced,4.694698651631673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,balanced,4.699808120727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,balanced,4.705343882242839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,balanced,4.721328099568685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,balanced,4.7391306559244795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,balanced,4.749317487080892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,balanced,4.77783997853597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,balanced,4.796426773071289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,balanced,4.814906756083171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,balanced,4.837973276774089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,balanced,4.917925198872884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,balanced,4.9588212966918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,balanced,5.05892276763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,balanced,5.199978510538737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,balanced,5.355440139770508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,balanced,5.634629567464192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,balanced,6.030053456624349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,balanced,6.456218719482422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,balanced,7.057200113932292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,balanced,9.532927831013998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,power_law_1.2,0.08393599987030029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,power_law_1.2,0.0841152012348175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,power_law_1.2,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,power_law_1.2,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,power_law_1.2,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,power_law_1.2,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,power_law_1.2,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,power_law_1.2,0.09043200016021728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,power_law_1.2,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,power_law_1.2,0.09254400134086609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,power_law_1.2,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,power_law_1.2,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,power_law_1.2,0.11523200273513794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,power_law_1.2,0.13612799644470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,power_law_1.2,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,power_law_1.2,0.1484544038772583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,power_law_1.2,0.19063680171966552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,power_law_1.2,0.27543039321899415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.2,0.33999359607696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.2,0.4687359809875488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.2,0.5503744125366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.2,0.8251199722290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.2,1.056441593170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.2,1.5188032150268556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.2,1.883839988708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.2,3.2711231231689455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,power_law_1.01,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,power_law_1.01,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,power_law_1.01,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,power_law_1.01,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,power_law_1.01,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,power_law_1.01,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,power_law_1.01,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,power_law_1.01,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,power_law_1.01,0.07157120108604431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,power_law_1.01,0.08728320002555848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,power_law_1.01,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,power_law_1.01,0.1601215958595276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,power_law_1.01,0.20466558933258056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,power_law_1.01,0.2700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,power_law_1.01,0.3347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,power_law_1.01,0.4750847816467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,power_law_1.01,0.6363584041595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.2,4.076057434082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.07492480278015137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.10883840322494506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.1338752031326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.14876799583435057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.1859328031539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.22542719841003417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.292902398109436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.39784319400787355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.5373631954193115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.6926271915435791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,power_law_1.01,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,power_law_1.01,0.028627198934555054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,power_law_1.01,0.030003198981285097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,power_law_1.01,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,power_law_1.01,0.04007680118083954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,power_law_1.01,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,power_law_1.01,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,power_law_1.01,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,power_law_1.01,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.08939520120620728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.1254591941833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.17447680234909058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.2141119956970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,power_law_1.01,0.2453632116317749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,power_law_1.2,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,power_law_1.2,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,power_law_1.2,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,power_law_1.2,0.12485760450363159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,power_law_1.2,0.1268288016319275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,power_law_1.2,0.13468159437179567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,power_law_1.2,0.14112000465393065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,power_law_1.2,0.1469696044921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,power_law_1.2,0.154310405254364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,power_law_1.2,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,power_law_1.2,0.1621440052986145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,power_law_1.2,0.16959359645843505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,power_law_1.2,0.163155198097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,power_law_1.2,0.18733439445495606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,power_law_1.2,0.19242240190505983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,power_law_1.2,0.2291775941848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,power_law_1.2,0.2522176027297974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,power_law_1.2,0.30951039791107177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,power_law_1.2,0.444704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,power_law_1.2,0.47409920692443847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,power_law_1.2,0.6848896026611329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,power_law_1.2,0.926790428161621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,balanced,0.07709866762161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,balanced,0.1070186694463094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,balanced,0.16102400422096252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,balanced,0.18583466609319052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,balanced,0.18833067019780478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,balanced,0.19201600551605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,balanced,0.19211200873057047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,balanced,0.19406400124231973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,balanced,0.19526400168736777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,balanced,0.19961599508921304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,balanced,0.1999946633974711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,balanced,0.20508799950281778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,balanced,0.21660800774892172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,balanced,0.21895466248194376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,balanced,0.22909333308537802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,balanced,0.2602826754252116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,balanced,0.2696426709493001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,power_law_1.2,1.4763903617858887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,balanced,0.33552531401316327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,balanced,0.36475733915964764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,balanced,0.5090239842732748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,balanced,0.5631999969482422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,balanced,0.7846773465474447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,balanced,0.9427200158437093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,balanced,1.4004640579223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,balanced,1.7451626459757488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.16833920478820802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.11306240558624267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.11195520162582398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.11880960464477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.1266111969947815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.13700480461120607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.14241280555725097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.16701439619064332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.22818560600280763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.2491136074066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.3094016075134277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.33578240871429443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,power_law_1.2,0.48059520721435545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,power_law_1.2,0.6154367923736572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,power_law_1.2,1.0026047706604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,power_law_1.2,1.096671962738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,power_law_1.2,1.8420223236083983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,power_law_1.2,2.2396799087524415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,power_law_1.2,3.322809600830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,balanced,0.0390079990029335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,balanced,0.0383146678407987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,balanced,0.04331733286380768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.06768533090750377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.08770133058230083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.09675199786822002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.12388267119725545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.15241066614786783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,balanced,0.07062399884064992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,balanced,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,balanced,0.08082133531570435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,balanced,0.09666132926940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,balanced,0.11308266719182332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,balanced,0.13184000054995218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,balanced,0.13402666648228964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,balanced,0.13052800297737122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,balanced,0.13473600149154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,balanced,0.13367467125256857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,balanced,0.13447999954223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,balanced,0.13675199945767721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,balanced,0.1377066671848297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,balanced,0.14754133423169455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,balanced,0.15067733327547708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,balanced,0.1542026698589325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,balanced,0.1629706621170044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,balanced,0.17097600301106772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,balanced,0.19260267416636148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,balanced,0.21083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,balanced,0.26039467255274457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,balanced,0.3070879975954692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,balanced,0.40979735056559247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,balanced,0.541045347849528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,balanced,0.7407733599344889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,balanced,0.9747040271759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.12089600563049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.16252800226211547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.22598400115966796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.22919039726257323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.23537919521331788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.2418368101119995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.23906559944152833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.25075199604034426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.24615039825439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.2560447931289673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.258790397644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.2682624101638794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.2754112005233765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.31370880603790285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.3220927953720093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.3171072006225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.35167999267578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.39147520065307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,power_law_1.01,0.4884352207183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,power_law_1.01,0.5589695930480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,power_law_1.01,0.7436031818389892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,power_law_1.01,0.9051584243774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,power_law_1.01,1.2789631843566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,power_law_1.2,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,power_law_1.2,0.029817599058151244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,power_law_1.2,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,power_law_1.2,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,power_law_1.2,0.033939200639724734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,power_law_1.01,1.6375167846679688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,power_law_1.2,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,power_law_1.01,2.374617576599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,power_law_1.2,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,power_law_1.2,0.18679039478302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,power_law_1.2,0.2506752014160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,power_law_1.2,0.3184063911437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,power_law_1.01,3.0744768142700196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,power_law_1.2,0.5002367973327637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,power_law_1.2,0.6094719886779785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,power_law_1.2,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.0682752013206482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,power_law_1.2,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.12172160148620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.13256959915161132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,power_law_1.2,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.18186240196228026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,power_law_1.2,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,power_law_1.2,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,power_law_1.2,0.09761279821395874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,power_law_1.2,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,power_law_1.2,0.14963200092315673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,power_law_1.2,0.21640961170196532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,power_law_1.2,0.23999359607696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,power_law_1.2,0.3559423923492432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,power_law_1.2,0.4540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.2571199893951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.33539841175079343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.5145664215087891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.0967296004295349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.09949439764022827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.10114560127258301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.1087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.12846720218658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.13127039670944213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.187827205657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.23223679065704345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.2790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,power_law_1.01,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.3754431962966919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,power_law_1.01,0.07351040244102477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.46111359596252444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,power_law_1.01,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,power_law_1.01,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,power_law_1.01,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,power_law_1.01,0.08115839958190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,power_law_1.01,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,power_law_1.01,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,power_law_1.01,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.01,0.11740800142288207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.01,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.01,0.1617792010307312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.01,0.2004863977432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.01,0.2517119884490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.01,0.3677504062652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.01,0.48363518714904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.01,0.6880512237548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.01,0.9687423706054688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,4,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,4,balanced,0.10269866387049358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,4,balanced,0.13778133193651834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,4,balanced,0.19828800360361734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,4,balanced,0.311898668607076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,4,balanced,0.3847413460413615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,4,balanced,0.38707200686136883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,4,balanced,0.39126400152842206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,4,balanced,0.3880639870961507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,4,balanced,0.3902346690495809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,4,balanced,0.3928639888763428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,4,balanced,0.39295466740926105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,4,balanced,0.39617598056793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,4,balanced,0.3993599812189738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,4,balanced,0.4044160048166911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,4,balanced,0.4066026608149211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,4,balanced,0.4125279982884725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,4,balanced,0.4238773187001546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,4,balanced,0.43689600626627606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,4,balanced,0.46727999051411945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,4,balanced,0.4934613307317098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,4,balanced,0.644810676574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,4,balanced,0.7094133694966634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,4,balanced,0.960256020228068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,4,balanced,1.2511093616485596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,4,balanced,1.626954714457194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.6428800106048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,4,balanced,2.1908532778422036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.8106047630310058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.1389951705932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.4869631767272948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,balanced,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,balanced,0.05442133545875549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,balanced,0.06452266871929169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,balanced,0.08522133032480876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,balanced,0.10273067156473796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,balanced,0.10629333058993022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,balanced,0.10428800185521443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,balanced,0.10453866918881734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,balanced,0.10579733053843181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,balanced,0.10749333103497823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,balanced,0.10900266965230306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,balanced,0.1090666651725769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,balanced,0.11825600266456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,balanced,0.12236799796422322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,balanced,0.13114133477210999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.1377226710319519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.15755732854207358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.17479467391967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.20708799362182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.2442773381868998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.3083626627922058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.3827413320541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.512389341990153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.6583573420842489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,balanced,0.049253334601720176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,balanced,0.07985599835713704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,balanced,0.09771733482678731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,balanced,0.1093280017375946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,balanced,0.1086293359597524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,balanced,0.10900800426801045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,balanced,0.10903466741243999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,balanced,0.10896000266075134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,balanced,0.11107732852300008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,balanced,0.11191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,balanced,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,balanced,0.11674132943153381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,balanced,0.12193600336710612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,balanced,0.12443199753761292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,balanced,0.127509335676829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,balanced,0.1381333371003469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,balanced,0.14896000425020853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,balanced,0.18703999121983847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,balanced,0.19825067122777304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,balanced,0.27883734305699664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,balanced,0.26070932547251385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,balanced,0.36347198486328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,balanced,0.41352001825968426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,balanced,0.5812586545944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,balanced,0.6918133099873861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,power_law_1.2,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,power_law_1.2,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,power_law_1.2,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.2,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.2,0.09244160056114196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.2,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.2,0.14689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.2,0.18356479406356813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.2,0.26513919830322263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.2,0.2907007932662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.2,0.4376704216003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,power_law_1.2,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,power_law_1.2,0.07704319953918456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,power_law_1.2,0.09260799884796142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,power_law_1.2,0.12844799757003783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,power_law_1.2,0.13498239517211913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,power_law_1.2,0.14306559562683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,power_law_1.2,0.14696320295333862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,power_law_1.2,0.1525823950767517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,power_law_1.2,0.1614848017692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,power_law_1.2,0.16943999528884887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,power_law_1.2,0.17939200401306152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,power_law_1.2,0.18063360452651978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,power_law_1.2,0.19173120260238646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,power_law_1.2,0.20242559909820557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,power_law_1.2,0.21885440349578858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,power_law_1.2,0.2625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,power_law_1.2,0.30981760025024413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,power_law_1.2,0.39141120910644533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,power_law_1.2,0.4660287857055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,power_law_1.2,0.619814395904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,power_law_1.2,0.8168512344360351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,power_law_1.2,1.1052224159240722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,power_law_1.2,1.4673791885375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,power_law_1.2,2.0439104080200194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,power_law_1.2,2.620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.2,0.6374015808105469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,power_law_1.2,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,power_law_1.2,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,power_law_1.2,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,power_law_1.2,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,power_law_1.2,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,power_law_1.2,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,power_law_1.2,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,power_law_1.2,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,power_law_1.2,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,power_law_1.2,0.11705600023269654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,power_law_1.2,0.13804160356521605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,power_law_1.2,0.17345919609069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,power_law_1.2,0.21546878814697265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,power_law_1.2,0.2781183958053589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,power_law_1.2,0.3669696092605591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,power_law_1.2,0.5207424163818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,power_law_1.2,0.6724160194396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,power_law_1.2,0.9886143684387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,power_law_1.2,1.2095808029174804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.01,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.01,0.0970687985420227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.01,0.10376319885253907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.01,0.21741440296173095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.01,0.21937921047210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.01,0.23299839496612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.01,0.24099199771881102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.01,0.24681599140167237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.01,0.2466495990753174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.01,0.25220479965209963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.01,0.2564480066299438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.01,0.2677056074142456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.01,0.27878398895263673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.01,0.28792319297790525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.01,0.30801920890808104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.01,0.36014719009399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.01,0.3855488061904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.01,0.48244481086730956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.01,0.5703296184539794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.01,0.719052791595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.01,0.8315199851989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.01,1.1364224433898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.01,1.3735615730285644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.01,1.9395263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.01,2.544595146179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,balanced,0.03125333289305369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,balanced,0.03179733455181122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,balanced,0.04229333500067393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,balanced,0.08500799536705017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,balanced,0.1048906644185384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,balanced,0.14140266180038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,balanced,0.1823306679725647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,balanced,0.2464053432146708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,balanced,0.3068959911664327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.10592000484466553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.11139199733734131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.11429760456085206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.11596800088882446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.11742080450057983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.11840640306472779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.12403839826583862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.12373759746551513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.12770559787750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.13134080171585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.13760639429092408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.1507904052734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.1661631941795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.20635519027709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.20999040603637695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.2585279941558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.32511999607086184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,power_law_1.01,0.3856384038925171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,power_law_1.01,0.47560958862304686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,power_law_1.01,0.6960127830505372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,power_law_1.01,0.9117695808410644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.08316159844398499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.08318079710006714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.10521600246429444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.10619519948959351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.11852799654006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.15820800065994262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.22024960517883302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,power_law_1.01,0.2633791923522949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,power_law_1.01,0.4371327877044678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,power_law_1.01,0.4704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,power_law_1.01,0.720966386795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,power_law_1.2,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,power_law_1.2,0.08438400030136109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,power_law_1.2,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,power_law_1.2,0.08506879806518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,power_law_1.2,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,power_law_1.2,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,power_law_1.2,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,power_law_1.2,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,power_law_1.2,0.09660159945487976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,power_law_1.2,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,power_law_1.2,0.10979199409484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,power_law_1.2,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,power_law_1.2,0.12061439752578736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,power_law_1.2,0.14058239459991456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.2,0.15556479692459108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.2,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.2,0.22634239196777345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.2,0.3043008089065552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.2,0.36792960166931155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.2,0.5985472202301025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.2,0.7900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.2,1.0888192176818847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,power_law_1.01,0.9201984405517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.2,1.7968255996704101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,balanced,0.050613333781560264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,balanced,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,balanced,0.12133866548538208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,balanced,0.12302933136622111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,balanced,0.12657599647839865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,balanced,0.1251573363939921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,balanced,0.12786666552225748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,balanced,0.12753066420555115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,balanced,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,balanced,0.13594667116800943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,balanced,0.13614400227864584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,balanced,0.14180800318717957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,balanced,0.1490133305390676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,balanced,0.1548853317896525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,balanced,0.16581867138544717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,balanced,0.18948266903559366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,balanced,0.21387199560801187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,balanced,0.27057067553202313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,balanced,0.31631465752919513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,balanced,0.47084800402323407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,balanced,0.5763680140177408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,balanced,0.8324053287506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,balanced,1.0665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,balanced,1.5787092844645183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,balanced,2.093407948811849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,power_law_1.01,0.07067520022392274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,power_law_1.01,0.08218880295753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,power_law_1.01,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,power_law_1.01,0.08970239758491516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,power_law_1.01,0.08856959939002991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,power_law_1.01,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,power_law_1.01,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,power_law_1.01,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,power_law_1.01,0.11099519729614257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,power_law_1.01,0.1251520037651062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.01,0.13918720483779906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.01,0.15774719715118407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.01,0.1888383984565735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.01,0.24622719287872313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.01,0.33120639324188234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.01,0.4543295860290527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.01,0.5864384174346924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.01,0.8632512092590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.01,1.1126144409179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.06088320016860962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.07480319738388061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.10917119979858399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.13112959861755372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.17569279670715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,power_law_1.01,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,power_law_1.01,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,power_law_1.01,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,power_law_1.01,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,power_law_1.01,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,power_law_1.01,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,power_law_1.01,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,power_law_1.01,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,power_law_1.01,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,power_law_1.01,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,power_law_1.01,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,power_law_1.01,0.10327039957046509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,power_law_1.01,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,power_law_1.01,0.15393919944763185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,power_law_1.01,0.19800959825515746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,power_law_1.01,0.2395263910293579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,power_law_1.01,0.33356800079345705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,power_law_1.01,0.42128639221191405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,power_law_1.01,0.618611192703247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,power_law_1.01,0.7638783931732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.21191039085388183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.3005759954452515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.47443199157714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,balanced,0.04641599953174591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,balanced,0.048101335763931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,balanced,0.046906664967536926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,balanced,0.09990933537483215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,balanced,0.1274773379166921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,balanced,0.1274666686852773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,balanced,0.1280586620171865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,balanced,0.12854400277137756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,balanced,0.133242666721344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,balanced,0.13357866803805032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,balanced,0.13505599896113077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,balanced,0.13674666484196982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,balanced,0.14140266180038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,balanced,0.1444906691710154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,balanced,0.15110400319099426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,balanced,0.16179733475049338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,balanced,0.17125332355499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,balanced,0.1951520045598348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,balanced,0.2142933408419291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,balanced,0.2587413390477498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,balanced,0.30665600299835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,balanced,0.3924959897994995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,balanced,0.4806506633758545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,balanced,0.6563359896341959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,balanced,0.8405386606852213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,power_law_1.01,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,power_law_1.01,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,power_law_1.01,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,power_law_1.01,0.11573760509490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,power_law_1.01,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,power_law_1.01,0.14469759464263915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,power_law_1.01,0.16665600538253783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,power_law_1.01,0.1668992042541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,power_law_1.01,0.17229440212249755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,power_law_1.01,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,power_law_1.01,0.1858944058418274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,power_law_1.01,0.1867136001586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,power_law_1.01,0.18929920196533204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,power_law_1.01,0.19690239429473877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,power_law_1.01,0.20901761054992676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,power_law_1.01,0.21470720767974855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,power_law_1.01,0.23681919574737548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.01,0.2608511924743652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.01,0.3204351902008057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.01,0.37118079662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.01,0.4782400131225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.01,0.5921472072601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.01,0.8209152221679688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.01,1.0091263771057128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.01,1.5379008293151855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.01,1.8671552658081054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.144652795791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.19653120040893554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.22677760124206542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.34408318996429443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.4482304096221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,balanced,0.0805866668621699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,balanced,0.17217600345611572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,balanced,0.28218666712443036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,balanced,0.28836800654729206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,balanced,0.2836693326632182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,balanced,0.2860853274663289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,balanced,0.28941865762074787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,balanced,0.2925013303756714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,balanced,0.30089600880940753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,balanced,0.30557866891225177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,balanced,0.314794659614563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,balanced,0.3261066675186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,balanced,0.33604268232981366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,balanced,0.36233067512512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,balanced,0.4107626676559448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,balanced,0.43966933091481525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,balanced,0.5194026629130045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,balanced,0.5874666770299276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,balanced,0.80021866162618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,balanced,0.9158293406168619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,balanced,1.3080906867980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,balanced,1.5901172955830891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,balanced,2.3092320760091147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,balanced,2.9506613413492837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,balanced,0.04655466477076212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,balanced,0.07547200222810109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,balanced,0.0981333355108897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,balanced,0.15434666474660239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,balanced,0.18337599436442056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,balanced,0.18452266852060953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,balanced,0.183296004931132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,balanced,0.18355733156204224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,balanced,0.1838080088297526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,balanced,0.18549867471059164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,balanced,0.19016534090042114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,balanced,0.18754667043685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,balanced,0.18993600209554037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,balanced,0.19862399498621622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,balanced,0.20173867543538412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,balanced,0.20535467068354288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,balanced,0.2196213404337565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,balanced,0.22642133633295694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,balanced,0.2728106578191121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,balanced,0.26774932940800983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,balanced,0.37467201550801593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,balanced,0.3702826499938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,balanced,0.531333327293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,balanced,0.5675893227259318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,balanced,0.8514506816864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,balanced,1.006112019220988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.15595519542694092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.2632575988769531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.46450557708740237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.7787968158721924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.0089344024658202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.6641023635864258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.7471935272216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,power_law_1.01,1.820863914489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,power_law_1.01,1.8749759674072266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,power_law_1.01,1.896780776977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,power_law_1.01,1.9509504318237305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.002207946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.047097587585449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.0975872039794923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.094643211364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.1117759704589845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.2065408706665037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,power_law_1.01,2.3867136001586915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,power_law_1.01,2.554099273681641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.008992004394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,power_law_1.01,3.122559928894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,power_law_1.01,3.7583297729492187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,power_law_1.01,3.7700225830078127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,power_law_1.01,4.607040023803711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,power_law_1.01,5.990988922119141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,power_law_1.01,7.635276794433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,power_law_1.01,8.363993835449218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,power_law_1.01,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,power_law_1.01,0.08535680174827576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,power_law_1.01,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,power_law_1.01,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,power_law_1.01,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,power_law_1.01,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,power_law_1.01,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,power_law_1.01,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,power_law_1.01,0.09079679846763611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,power_law_1.01,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,power_law_1.01,0.1009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,power_law_1.01,0.11854720115661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,power_law_1.01,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.01,0.12895359992980956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.01,0.1551424026489258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.01,0.17838079929351808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.01,0.22984960079193115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.01,0.28152320384979246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.01,0.3738303899765015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.01,0.49367680549621584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.01,0.7217023849487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.01,0.9372544288635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.01,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.01,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.01,0.047391998767852786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.01,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.01,0.06993920207023621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.01,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.01,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.01,0.12433919906616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.01,0.16108800172805787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.01,0.20275840759277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.01,0.2458048105239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,balanced,0.03514133393764496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,balanced,0.03591466695070267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,balanced,0.0658186674118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,balanced,0.06482133269309998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,balanced,0.08220799763997395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,balanced,0.10652800401051839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,balanced,0.11339199542999268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.2,0.1276479959487915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.2,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.2,0.2754175901412964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.2,0.37022719383239744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.2,0.6390016078948975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.2,0.6870463848114013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.2,0.7189888000488281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.2,0.7437759876251221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.2,0.750713586807251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.2,0.7836415767669678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.2,0.794700813293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.2,0.8368831634521484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.2,0.8492032051086426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.2,0.8799360275268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.2,0.902137565612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.2,0.9348031997680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.2,1.0294272422790527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.2,1.1131648063659667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.2,1.282963180541992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.2,1.3885439872741698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.2,1.7046911239624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.10366079807281495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.135916805267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.2,1.8985471725463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.18787200450897218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.2910655975341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.4055039882659912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.2,2.453318405151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.5193471908569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.7531328201293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.2,3.1465471267700194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7913919925689697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8156224250793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8610943794250489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8843520164489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.902284812927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.9286335945129395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.925011157989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.2,4.419059371948242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.9910783767700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,0.9881664276123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.0552448272705077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.119270420074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.2066816329956054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.3550592422485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.4426752090454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,1.6922687530517577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,1.9691648483276367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,2.5188671112060548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.2,5.3317310333251955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,2.962924766540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,3.9999935150146486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,5.060710525512695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,balanced,0.11455466349919637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,balanced,0.1186133325099945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,balanced,0.11199999849001567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,balanced,0.11574400464693706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,balanced,0.11524800459543864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,balanced,0.1150933305422465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,balanced,0.1183626651763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,balanced,0.11973866820335388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,balanced,0.12628799676895142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,balanced,0.12643733620643616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,balanced,0.13074666261672974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,balanced,0.13753599921862283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,balanced,0.14507733782132468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,balanced,0.16346133748690286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,balanced,0.1772480010986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,balanced,0.21294933557510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,balanced,0.2422986626625061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,balanced,0.32261866331100464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,balanced,0.3744800090789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,balanced,0.5234026511510214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,balanced,0.6485653320948283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,power_law_1.01,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,power_law_1.01,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,power_law_1.01,0.07937920093536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,power_law_1.01,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,power_law_1.01,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,power_law_1.01,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,power_law_1.01,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,power_law_1.01,0.0887935996055603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,power_law_1.01,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,power_law_1.01,0.10440319776535034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,power_law_1.01,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,power_law_1.01,0.11167360544204712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,power_law_1.01,0.1301695942878723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,power_law_1.01,0.1482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,power_law_1.01,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,power_law_1.01,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,power_law_1.01,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,power_law_1.01,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,power_law_1.01,0.18808319568634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,power_law_1.01,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,power_law_1.01,0.08828160166740417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,power_law_1.01,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,power_law_1.01,0.09575679898262024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,power_law_1.01,0.09834240078926086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,power_law_1.01,0.10707839727401733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,power_law_1.01,0.11884160041809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,power_law_1.01,0.12951680421829223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,power_law_1.01,0.1497663974761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,power_law_1.01,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,power_law_1.01,0.20697600841522218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,power_law_1.01,0.24513280391693115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,power_law_1.01,0.3326080083847046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,power_law_1.01,0.4184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,power_law_1.01,0.5931327819824219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,power_law_1.01,0.7612671852111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,power_law_1.01,1.0952511787414552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,power_law_1.01,0.22842879295349122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,power_law_1.01,1.4318911552429199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,power_law_1.01,0.3046463966369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,power_law_1.01,0.3607743978500366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,power_law_1.01,0.5118207931518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,power_law_1.01,0.6416575908660889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,power_law_1.01,0.9069696426391601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,power_law_1.01,1.1819007873535157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,power_law_1.01,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,power_law_1.01,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,power_law_1.01,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,power_law_1.01,0.05546879768371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,power_law_1.01,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,power_law_1.01,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.01,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.01,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.01,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.01,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.01,0.1722815990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.01,0.23204479217529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.01,0.2771519899368286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.01,0.4146880149841309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.01,0.5065023899078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,balanced,0.029706666866938274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,balanced,0.029781334102153778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,balanced,0.029765332738558452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,balanced,0.02770666778087616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,balanced,0.03195200115442276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,balanced,0.032442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,balanced,0.03189333279927572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,balanced,0.03480533262093862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,balanced,0.036831999818483986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,balanced,0.11741866668065389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,balanced,0.14457600315411887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,power_law_1.01,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,power_law_1.01,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,power_law_1.01,0.030720001459121703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,power_law_1.01,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,power_law_1.01,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,power_law_1.01,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,power_law_1.01,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,power_law_1.01,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,power_law_1.01,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,power_law_1.01,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,power_law_1.01,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,balanced,0.05379733443260193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,balanced,0.0765226682027181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,balanced,0.07866133252779643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,balanced,0.0798933357000351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,balanced,0.07842666904131572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,balanced,0.08281599978605907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,balanced,0.09075733025868733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,balanced,0.09112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,balanced,0.09699733058611552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,balanced,0.10544000069300334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,balanced,0.11260799566904704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,balanced,0.1360213359196981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,balanced,0.156442662080129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,balanced,0.19155200322469076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,balanced,0.22394132614135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,balanced,0.3161226709683736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,balanced,0.3890293439229329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,balanced,0.5603626569112142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,balanced,0.7401866912841797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,power_law_1.01,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,power_law_1.01,0.1751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,power_law_1.01,0.24461441040039061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,power_law_1.01,0.29999361038208006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,power_law_1.01,0.4434815883636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,power_law_1.01,0.5741951942443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.0743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.10602240562438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.12539520263671874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.1386944055557251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.17633919715881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.1899392008781433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.22544000148773194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.34038400650024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.40789117813110354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.6638527870178222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,0.9415103912353515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.1963839530944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,1.60263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,power_law_1.01,0.082150399684906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,power_law_1.01,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,power_law_1.01,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,power_law_1.01,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,power_law_1.01,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,power_law_1.01,0.06805760264396668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,power_law_1.01,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,power_law_1.01,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,power_law_1.01,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,power_law_1.01,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,power_law_1.01,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,power_law_1.01,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,power_law_1.01,0.10516480207443238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,power_law_1.01,0.11697280406951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,power_law_1.01,0.1462847948074341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,power_law_1.01,0.16072319746017455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,power_law_1.01,0.20164480209350585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,power_law_1.01,0.25253119468688967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,power_law_1.01,0.3329024076461792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,power_law_1.01,0.43401598930358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,power_law_1.01,0.6184576034545899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,power_law_1.01,0.7901440143585206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.14922239780426025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.1862015962600708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.18075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.19434239864349365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.1955456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.1625216007232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.1701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.15920640230178834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.1745792031288147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.1611583948135376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.18435839414596558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.18773759603500367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.1937343955039978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.18369280099868773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.19036799669265747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.20758399963378907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.22392959594726564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.2434175968170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.28929920196533204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.30580480098724366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.39266560077667234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.38496639728546145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,power_law_1.2,0.5270656108856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,power_law_1.2,0.6040383815765381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,power_law_1.2,0.9190143585205078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,power_law_1.2,1.1195008277893066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.0732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.09041280150413514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.10594559907913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.15383679866790773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.22081279754638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.31008639335632326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,power_law_1.2,0.438259220123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,power_law_1.2,0.5493567943572998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.01,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.01,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.01,0.02898559868335724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.01,0.029395198822021483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.01,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.01,0.03075839877128601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.01,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.01,0.18618240356445312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.01,0.22672638893127442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,1,power_law_1.2,0.12242560386657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,1,power_law_1.2,0.13834240436553955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,1,power_law_1.2,0.16726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,1,power_law_1.2,0.19639040231704713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,1,power_law_1.2,0.22584960460662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,1,power_law_1.2,0.2605247974395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,1,power_law_1.2,0.33476479053497316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,1,power_law_1.2,0.34590721130371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,1,power_law_1.2,0.3641535997390747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,1,power_law_1.2,0.3665087938308716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,1,power_law_1.2,0.38010239601135254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,1,power_law_1.2,0.38569600582122804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,1,power_law_1.2,0.3956671953201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,1,power_law_1.2,0.4119679927825928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,1,power_law_1.2,0.4111167907714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,1,power_law_1.2,0.4251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,1,power_law_1.2,0.44467840194702146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,1,power_law_1.2,0.4702847957611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.2,0.5085504055023193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.2,0.5895999908447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.2,0.6658944129943848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.2,0.7944831848144531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.2,0.9670016288757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.2,1.2810367584228515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.2,1.6068031311035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.2,2.2497215270996094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.2,2.875494384765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,power_law_1.2,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,power_law_1.2,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,power_law_1.2,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,power_law_1.2,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,power_law_1.2,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,power_law_1.2,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,power_law_1.2,0.12795519828796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,power_law_1.2,0.15576319694519042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,power_law_1.2,0.2168895959854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,power_law_1.2,0.2893887996673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.01,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.01,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.01,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.01,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.01,0.06853119730949402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.01,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.01,0.10877439975738526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.01,0.13110400438308717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.01,0.1614400029182434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.01,0.20113921165466309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.01,0.23849599361419677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.01,0.32503681182861327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.01,0.4216896057128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,power_law_1.2,0.36650240421295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,power_law_1.2,0.5569664001464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.2,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.2,0.09450240135192871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.2,0.127347195148468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.2,0.16821119785308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.2,0.21777279376983644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.2,0.35438079833984376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.2,0.3874495983123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.2,0.3937407970428467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.2,0.41189122200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.2,0.41716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.2,0.4321599960327148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.2,0.449235200881958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.2,0.45981440544128416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.2,0.4847551822662354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.2,0.4995584011077881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.2,0.5163136005401612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.2,0.5518847942352295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.2,0.6444736003875733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.2,0.6817535877227783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.2,0.8439167976379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.2,0.9944255828857422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.2,1.256230354309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.2,1.4689023971557618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.2,1.9944448471069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.2,2.4128128051757813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.2,3.3715648651123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.2,4.443859100341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.2,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.2,0.10397440195083618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.2,0.11512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.2,0.1106495976448059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.2,0.11897599697113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.2,0.11816960573196411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.2,0.12336000204086303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.2,0.12579200267791749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.2,0.1268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.2,0.1273151993751526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.2,0.13579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.2,0.14321279525756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.2,0.14634239673614502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.2,0.16732159852981568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.2,0.16990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.2,0.1997439980506897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.2,0.21060481071472167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.2,0.2606911897659302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.2,0.2835520029067993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.2,0.3641599893569946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.2,0.449721622467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.2,0.6247615814208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,power_law_1.2,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,power_law_1.2,0.08680319786071777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,power_law_1.2,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,power_law_1.2,0.11414400339126587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,power_law_1.2,0.11971839666366577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,power_law_1.2,0.13915519714355468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,power_law_1.2,0.14065279960632324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,power_law_1.2,0.15998719930648803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,power_law_1.2,0.15911680459976196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,power_law_1.2,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.2,0.784607982635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,power_law_1.2,0.15622400045394896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,power_law_1.2,0.16714240312576295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,power_law_1.2,0.1627071976661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,power_law_1.2,0.17166080474853515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,power_law_1.2,0.17900160551071168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,power_law_1.2,0.18375680446624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,power_law_1.2,0.20102400779724122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.2,0.22766079902648925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.2,0.2645695924758911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.2,0.32863359451293944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.2,0.42583041191101073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.2,0.5118847846984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.2,0.7161920070648193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.2,0.9146047592163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.2,1.3599167823791505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.2,1.8268800735473634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,balanced,0.046122665206591286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,balanced,0.04284800092379252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,balanced,0.05052266518274943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,balanced,0.04756266872088114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,balanced,0.06458133459091187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,balanced,0.07217066486676534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,balanced,0.08434666196505229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,balanced,0.09493866562843323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,balanced,0.12191466490427653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,balanced,0.14827199776967367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,balanced,0.20107734203338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,balanced,0.2536960045496623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,balanced,0.04429866870244344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,balanced,0.10249066352844238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,balanced,0.12812266747156778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,balanced,0.13160000244776407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,balanced,0.12939199805259705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,balanced,0.13052800297737122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,balanced,0.12993066509564719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,balanced,0.13194666306177774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,balanced,0.13426666458447775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,balanced,0.13242133458455405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,balanced,0.13362666964530945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,balanced,0.14014400045077005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,balanced,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,balanced,0.1437279979387919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,balanced,0.1567039986451467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,balanced,0.1569546659787496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,balanced,0.18831467628479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,balanced,0.18203733364741007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,balanced,0.24645866950352988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,balanced,0.24986666440963745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,balanced,0.34590399265289307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,balanced,0.3807520071665446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,balanced,0.5507306655248007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,balanced,0.6363146702448527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,balanced,0.048197334011395775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,balanced,0.053685332338015236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,balanced,0.0558240016301473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,balanced,0.05694933235645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,balanced,0.061861331264177956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.07062399884064992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.1256160040696462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.14098667105038962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.17599999904632568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.2031466762224833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.28593599796295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.35050666332244873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,power_law_1.2,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,power_law_1.2,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,power_law_1.2,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,power_law_1.2,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,power_law_1.2,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.2,0.11055359840393067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.2,0.13866879940032958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.2,0.1908735990524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.2,0.2967103958129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.2,0.35765120983123777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.2,0.519481611251831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.2,0.7144576072692871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.08051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.08564479947090149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.09872000217437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.11818239688873292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.13749120235443116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.1781823992729187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.2143359899520874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.29499518871307373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.37547519207000735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,balanced,0.051354666550954185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,balanced,0.07561066746711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,balanced,0.12026133139928182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,balanced,0.11915733416875203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,balanced,0.12033067146937053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,balanced,0.11844799915949504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,balanced,0.1199626624584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,balanced,0.12129066387812297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,balanced,0.12155733505884807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,balanced,0.12370133399963379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,balanced,0.1237386663754781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,balanced,0.12411200006802876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,balanced,0.13170133034388223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,balanced,0.1317813297112783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,balanced,0.1376106639703115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,balanced,0.14222400387128195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,balanced,0.15001599987347922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,balanced,0.17562133073806763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,balanced,0.18560532728830972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,balanced,0.22872533400853476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,balanced,0.2586560050646464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,balanced,0.36130134264628094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,balanced,0.40957868099212646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,balanced,0.5676266749699911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.5385151863098144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,balanced,0.721226692199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6957824230194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.09588479995727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.12999039888381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.15425280332565308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.17297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.14923520088195802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,power_law_1.01,0.02563839852809906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.16142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,power_law_1.01,0.027321600914001466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,power_law_1.01,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.16727039813995362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,power_law_1.01,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.17471359968185424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.17788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,power_law_1.01,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.17729920148849487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,power_law_1.01,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.18261120319366456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,power_law_1.01,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,power_law_1.01,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.19022079706192016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,power_law_1.01,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.21084799766540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,power_law_1.01,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.2300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.25131518840789796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.2858815908432007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.30732800960540774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.3853247880935669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,power_law_1.01,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.4158912181854248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,power_law_1.2,0.5354176044464112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.13245439529418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,power_law_1.2,0.6714176177978516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.19434239864349365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,power_law_1.2,0.9706815719604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.2573632001876831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.33362560272216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,power_law_1.2,1.132703971862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,power_law_1.01,0.432908821105957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,balanced,0.03938133269548416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,balanced,0.04411733150482178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,balanced,0.044112001856168113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,balanced,0.05119466781616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,balanced,0.057002668579419456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,balanced,0.06131199995676676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,balanced,0.07253866891066234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,balanced,0.09692266583442688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,balanced,0.12568533420562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,balanced,0.15421866377194723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,balanced,0.21174399058024088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,balanced,0.26146133740743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,power_law_1.01,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,power_law_1.01,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,power_law_1.01,0.08707839846611024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,power_law_1.01,0.10301439762115479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,power_law_1.01,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,power_law_1.01,0.11455359458923339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,power_law_1.01,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,power_law_1.01,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,power_law_1.01,0.12269439697265624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,power_law_1.01,0.13119360208511352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,power_law_1.01,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,power_law_1.01,0.13863680362701417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,power_law_1.01,0.17447680234909058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,power_law_1.01,0.18959360122680663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,power_law_1.01,0.2344383955001831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,power_law_1.01,0.279801607131958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,power_law_1.01,0.35525760650634763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,power_law_1.01,0.4326144218444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,power_law_1.01,0.5896255970001221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,power_law_1.01,0.7385983943939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,power_law_1.01,1.0575615882873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,power_law_1.01,1.3660287857055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.1527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.19541759490966798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.24632320404052735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.3295232057571411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.4003392219543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,power_law_1.2,0.5602431774139405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,power_law_1.2,0.7774208068847657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,power_law_1.01,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,power_law_1.01,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,power_law_1.01,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,power_law_1.01,0.04586879909038544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,power_law_1.01,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,power_law_1.01,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,power_law_1.01,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,power_law_1.01,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,power_law_1.01,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,power_law_1.01,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,power_law_1.01,0.11828479766845704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,power_law_1.01,0.18161920309066773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,power_law_1.01,0.2374336004257202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,power_law_1.01,0.32906880378723147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,power_law_1.01,0.4175104141235352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,power_law_1.01,0.5929664134979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,power_law_1.01,0.8454143524169921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,balanced,0.056405335664749146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,balanced,0.054458667834599815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,balanced,0.06091733276844025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,balanced,0.06411733229955037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,balanced,0.07156266768773396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,balanced,0.08265600105126698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.08674133817354839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.10521066188812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.11934933066368103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.1511840025583903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.18077333768208823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.23340266942977905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.29386667410532635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.4036159912745158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.5146346489588419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.01,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.01,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.01,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.01,0.04953599870204926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.01,0.11242879629135132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.01,0.12231040000915527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.01,0.1713152050971985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.01,0.20461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.0883072018623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.13420799970626832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.1609344005584717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.1891584038734436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.2425856113433838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.2957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.4052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5142399787902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.7250304222106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,2,power_law_1.2,0.11012480258941651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,2,power_law_1.2,0.13259520530700683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,2,power_law_1.2,0.15978879928588868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,2,power_law_1.2,0.2081984043121338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,2,power_law_1.2,0.23422720432281494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,2,power_law_1.2,0.33218560218811033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,2,power_law_1.2,0.3410943984985352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,balanced,0.07744533320267995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,balanced,0.11158399780591328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,2,power_law_1.2,0.3667903900146484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,balanced,0.1750719944636027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,balanced,0.20565332969029745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,balanced,0.2135093410809835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,2,power_law_1.2,0.3614784002304077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,balanced,0.21486934026082358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,balanced,0.21570666631062826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,balanced,0.2188106576601664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,balanced,0.2232159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,2,power_law_1.2,0.3794879913330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,balanced,0.2262293299039205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,balanced,0.23027199506759644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,balanced,0.23894399404525757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,2,power_law_1.2,0.395251202583313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,balanced,0.24452267090479532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,balanced,0.2587040066719055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,balanced,0.2725119988123576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,balanced,0.31204267342885333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,2,power_law_1.2,0.39774720668792723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,balanced,0.33555734157562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,balanced,0.44655466079711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,balanced,0.48417067527770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,2,power_law_1.2,0.3985088109970093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,balanced,0.7060266335805258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,balanced,0.796773354212443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,2,power_law_1.2,0.4111040115356445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,balanced,1.2611786524454753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,2,power_law_1.2,0.4117760181427002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,balanced,1.443130652109782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,2,power_law_1.2,0.4164415836334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,balanced,2.170687993367513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,2,power_law_1.2,0.43232002258300783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,2,power_law_1.2,0.47273597717285154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,balanced,2.729951858520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.2,0.5206208229064941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.2,0.5956223964691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.2,0.7056640148162842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.2,0.9189632415771485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.2,1.1714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.2,1.5070591926574708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.2,1.9747648239135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.2,2.7135808944702147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.9281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.2,3.640959930419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,balanced,0.08049599826335907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,balanced,0.0842026670773824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,balanced,0.08379200100898743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,balanced,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,balanced,0.08668800195058186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,balanced,0.08675733208656311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,balanced,0.08864532907803853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,balanced,0.08842133482297261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,balanced,0.08867200215657552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,balanced,0.09267733494440715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,balanced,0.09230933586756389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,balanced,0.09518399834632874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,balanced,0.10677867134412129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,balanced,0.10745066404342651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,balanced,0.11053333679835002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,balanced,0.13015466928482056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,balanced,0.1398293375968933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,balanced,0.17069333791732788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,balanced,0.19688532749811807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,balanced,0.2647413412729899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,balanced,0.3301279942194621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,balanced,0.46429868539174396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,balanced,0.5955359935760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,balanced,0.8612319628397623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,balanced,1.1167199611663818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.09980159997940063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.14881919622421264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.22118399143218995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.3220607995986938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.4321023941040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.5934463977813721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.6026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,power_law_1.2,0.8555968284606934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,power_law_1.2,0.8640128135681152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,power_law_1.2,0.8985152244567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,power_law_1.2,0.9220095634460449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,power_law_1.2,0.974176025390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,power_law_1.2,0.9900927543640137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,power_law_1.2,1.0034624099731446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.0453887939453126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.0393407821655274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.0698368072509765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.154252815246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.2651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.4515904426574706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,power_law_1.2,1.4623295783996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,power_law_1.2,1.7320383071899415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,power_law_1.2,1.9441280364990234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,power_law_1.2,2.462924766540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,power_law_1.2,2.45599365234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,power_law_1.2,2.990892791748047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,power_law_1.2,3.8342849731445314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,power_law_1.2,0.06032000184059143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,power_law_1.2,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,power_law_1.2,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,power_law_1.2,0.08989440202713013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,power_law_1.2,0.10399999618530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,power_law_1.2,0.10948480367660522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,power_law_1.2,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,power_law_1.2,0.11489919424057007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,power_law_1.2,0.11487360000610351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,power_law_1.2,0.12257280349731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,power_law_1.2,0.12911360263824462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,power_law_1.2,0.13292160034179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,power_law_1.2,0.14130560159683228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,power_law_1.2,0.15964800119400024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.2,0.17377279996871947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.2,0.21406080722808837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.2,0.2500927925109863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.2,0.32284159660339357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.2,0.40977277755737307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.2,0.5518335819244384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.2,0.7129343986511231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.2,1.005964756011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.2,1.3418047904968262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,balanced,0.04099733382463455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,balanced,0.04200533529122671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,balanced,0.04667733112970988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,balanced,0.06037333110968272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,balanced,0.0666720022757848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,balanced,0.07065066695213318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,balanced,0.07259200016657512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,balanced,0.07382399837176006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,balanced,0.07448533177375793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,balanced,0.07753066718578339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,balanced,0.08340266346931458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,balanced,0.08544533451398213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,balanced,0.09285866220792134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,balanced,0.11125866572062175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,balanced,0.12153599659601848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,balanced,0.1483840048313141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,balanced,0.17108800013860068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,balanced,0.21535466114679971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,balanced,0.277349332968394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,balanced,0.3736693461736043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,balanced,0.485365351041158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,balanced,0.6993227005004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,balanced,0.8919626871744791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,power_law_1.01,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,power_law_1.01,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,power_law_1.01,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,power_law_1.01,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,power_law_1.01,0.0531328022480011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,power_law_1.01,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.01,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.01,0.10609920024871826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.01,0.13995519876480103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.01,0.18030719757080077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.01,0.2604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.01,0.3505728006362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.01,0.47640318870544435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.01,0.7090816020965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.01,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.01,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.01,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.01,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.01,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.01,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.01,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.01,0.07585920095443725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.01,0.10033279657363892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.01,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.01,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.01,0.18554879426956178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.01,0.18284159898757935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.01,0.24332799911499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.01,0.2878592014312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.01,0.36681599617004396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.01,0.4492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,power_law_1.2,0.10106879472732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,power_law_1.2,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,power_law_1.2,0.19016319513320923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,power_law_1.2,0.192576003074646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,power_law_1.2,0.19233920574188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,power_law_1.2,0.1935871958732605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,power_law_1.2,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,power_law_1.2,0.17611520290374755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,power_law_1.2,0.18305280208587646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,power_law_1.2,0.16962560415267944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,power_law_1.2,0.16959999799728392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,power_law_1.2,0.1855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,power_law_1.2,0.18260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,power_law_1.2,0.18076159954071044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,power_law_1.2,0.18565759658813477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,power_law_1.2,0.1966271996498108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,power_law_1.2,0.21719679832458497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,power_law_1.2,0.23865599632263185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,power_law_1.2,0.25742719173431394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,power_law_1.2,0.3171776056289673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,power_law_1.2,0.349183988571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,power_law_1.2,0.43405442237854003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,power_law_1.2,0.5104896068572998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,power_law_1.2,0.6952576160430908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,power_law_1.2,0.7613056182861329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,power_law_1.2,1.0773887634277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,power_law_1.2,1.363699245452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.11543040275573731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.14529279470443726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.1513983964920044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.15203839540481567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.1589184045791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.1636415958404541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.16908799409866332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.17351679801940917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.18426239490509033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.18780800104141235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.1985535979270935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.222489595413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.2475264072418213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.3037503957748413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.3403072118759155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.44515199661254884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.5536511898040771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.7385663986206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,1.0120448112487792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.3766271591186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,1.7839359283447265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.07488639950752259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.08778240084648133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.11440000534057618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.1335935950279236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.15241600275039674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.19124480485916137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.21676158905029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.30037760734558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.36884479522705077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.526796817779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.677836799621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.07537919878959656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.093094402551651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.12969599962234496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.16692479848861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.1814463973045349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.22552320957183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.3013695955276489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.3828416109085083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.4901567935943604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,power_law_1.2,0.6786880016326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,power_law_1.2,0.9087039947509765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,balanced,0.06632533172766368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,balanced,0.08861866593360901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,balanced,0.08668800195058186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,balanced,0.09301867087682088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,balanced,0.09489066402117412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,balanced,0.09691733121871948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,balanced,0.10063466429710388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,balanced,0.10942932963371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,balanced,0.11383466919263203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,balanced,0.12355200449625652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,balanced,0.14849066734313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,balanced,0.16850666205088297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,balanced,0.20775999625523886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,balanced,0.24716800451278687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,balanced,0.35922133922576904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,balanced,0.4458773136138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,balanced,0.6507519880930582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,balanced,0.844048023223877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,balanced,1.2577333450317383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,balanced,1.65338134765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,balanced,0.03688533355792364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,balanced,0.036789332826932274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,balanced,0.03730666637420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,balanced,0.04252799848715464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,balanced,0.042725334564844765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,balanced,0.04344533383846283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,balanced,0.045498669147491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,balanced,0.06163200239340464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,balanced,0.0660693347454071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,balanced,0.10285333792368571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,balanced,0.1413706640402476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,balanced,0.16883732875188193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,balanced,0.060234665870666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,balanced,0.057802667220433555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,balanced,0.0683786670366923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,balanced,0.06991466879844666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,balanced,0.07236266632874806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,balanced,0.07214933137098949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,balanced,0.0754559983809789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,balanced,0.07920533418655396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,balanced,0.08463467160860698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,balanced,0.08307733138402303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,balanced,0.09993599851926167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,balanced,0.10712533195813496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,balanced,0.12972266475359598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,balanced,0.1525226632754008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,balanced,0.19112000862757364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,balanced,0.18712000052134195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,balanced,0.24257065852483115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,balanced,0.2643786668777466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,balanced,0.3550666570663452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,balanced,0.4183359940846761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,balanced,0.04516266783078512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,balanced,0.07725866635640462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,balanced,0.10264533758163452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,balanced,0.11574932932853699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,balanced,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,balanced,0.11691733201344807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,balanced,0.11902399857838948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,balanced,0.12148267030715942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,balanced,0.12031466762224834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,balanced,0.12242133418718974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,balanced,0.12274666627248128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,balanced,0.12628266215324402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,balanced,0.12964800000190735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,balanced,0.13499733805656433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,balanced,0.13763733704884848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,balanced,0.15335466464360556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,balanced,0.15922666589419046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,balanced,0.19105066855748495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,balanced,0.20379199584325156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,balanced,0.27850133180618286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,balanced,0.29741867383321124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,balanced,0.4305066665013631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,balanced,0.4985920190811157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,balanced,0.6702346801757812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,balanced,0.8723733425140381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,power_law_1.2,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,power_law_1.2,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,power_law_1.2,0.24702720642089843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,power_law_1.2,0.4391295909881592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,power_law_1.2,0.6841728210449218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,power_law_1.2,0.9429056167602539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,power_law_1.2,1.1468735694885255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,power_law_1.2,1.5597248077392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,power_law_1.2,1.614303970336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,power_law_1.2,1.7123136520385742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,power_law_1.2,1.793452835083008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,power_law_1.2,1.8427391052246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,power_law_1.2,1.892153549194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,power_law_1.2,1.9395391464233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,power_law_1.2,2.000595283508301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,power_law_1.2,2.0214080810546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,power_law_1.2,2.149337577819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,power_law_1.2,2.297542381286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,power_law_1.2,2.467571258544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,power_law_1.2,2.735193634033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,power_law_1.2,2.809113693237305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,power_law_1.2,3.3734462738037108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,power_law_1.2,3.4537662506103515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,power_law_1.2,4.252447891235351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,power_law_1.2,4.746758270263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.15991040468215942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.21644160747528077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.3461888074874878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.4310463905334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.4248191833496094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,power_law_1.2,5.9297534942626955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.44248318672180176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.4414656162261963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.44208641052246095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.44279041290283205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.44318079948425293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.4453375816345215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,power_law_1.2,6.98919677734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.45253758430480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.45586562156677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.47343997955322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.511027193069458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.5293375968933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.5674047946929932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.6827839851379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,0.75731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,0.9969023704528809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.2136575698852539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,1.683635139465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,2.1722368240356444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,3.1213823318481446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,4.080633544921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,6.008787155151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,7.8404541015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.11008640527725219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.11482880115509034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.11490559577941895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.12165759801864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.1285696029663086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.12487679719924927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.1394752025604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.13983999490737914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.14597760438919066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.16091519594192505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.17681280374526978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.21685760021209716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.2367232084274292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.30538880825042725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.37156479358673095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.47829761505126955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,power_law_1.01,0.6104320049285888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,power_law_1.01,0.8552960395812989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,power_law_1.01,1.0714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,balanced,0.04897599915663401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,balanced,0.06687999765078227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,balanced,0.06980800131956737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,balanced,0.07417066891988118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,balanced,0.10429333647092183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,balanced,0.11156800389289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,balanced,0.14018666744232178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,balanced,0.17296000321706137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,balanced,0.2368533412615458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,balanced,0.30637333790461224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,balanced,0.44147733847300213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,balanced,0.5817919969558716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,balanced,0.8339520295461019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,balanced,1.086026668548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,power_law_1.01,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,power_law_1.01,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,power_law_1.01,0.15018880367279053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,power_law_1.01,0.17548160552978515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,power_law_1.01,0.19984639883041383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,power_law_1.01,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,power_law_1.01,0.21575679779052734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,power_law_1.01,0.22398080825805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,power_law_1.01,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,power_law_1.01,0.2286463975906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,power_law_1.01,0.23639039993286132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,power_law_1.01,0.23838720321655274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,power_law_1.01,0.2500607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,power_law_1.01,0.26289279460906984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,power_law_1.01,0.2717375993728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,power_law_1.01,0.3136512041091919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,power_law_1.01,0.3114559888839722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,power_law_1.01,0.3703104019165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,power_law_1.01,0.4112832069396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,power_law_1.01,0.4942016124725342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,power_law_1.01,0.594432020187378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,power_law_1.01,0.7892672061920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,power_law_1.01,0.9773823738098144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,power_law_1.01,1.2445055961608886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,power_law_1.01,1.6318016052246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.0805184006690979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.11633280515670777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.12698880434036255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.129203200340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.13237119913101197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.13780479431152343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.13854080438613892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.1466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.15381120443344115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.16240639686584474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.16779520511627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.18316160440444945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.20921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.2323904037475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.28743040561676025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.3260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.42467198371887205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.5170944213867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.7126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,power_law_1.2,0.8888640403747559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.2838144302368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.6509567260742188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.11183359622955322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.13344000577926635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.17203840017318725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.1954624056816101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.19976320266723632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.19422080516815185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.1991744041442871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.20241920948028563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.2083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.2121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.21765758991241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.21504640579223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.22773759365081786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.2336064100265503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.24560000896453857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.2736255884170532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.2964672088623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.36126720905303955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.394214391708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.49923200607299806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.6107200145721435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.7578815937042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,0.9196991920471191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.3177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,1.6999168395996094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.2,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.2,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.2,0.09896320104598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.2,0.09644160270690919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.2,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.2,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.2,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.2,0.10502400398254394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.2,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.2,0.10917760133743286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.2,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.2,0.11961599588394164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.2,0.12076799869537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.2,0.14680960178375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.2,0.15909759998321532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.2,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.2,0.2441728115081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.2,0.28933119773864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.2,0.3278143882751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.2,0.43992958068847654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.2,0.5135488033294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.2,0.7190656185150146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.2,0.8474111557006836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,balanced,0.06964799761772156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,balanced,0.0513866643110911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,balanced,0.06618666648864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.09108266234397888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.10708266496658325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.11934399604797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.14427733421325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.17070933183034262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.21946134169896445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.26495466629664105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.3668160041173299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.46688000361124676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,balanced,0.06440000236034393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,balanced,0.06435733536879222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,balanced,0.06738666693369548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,balanced,0.0660693347454071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,balanced,0.07083199918270111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,balanced,0.07105599840482076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,balanced,0.07351466516653697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,balanced,0.07765333354473114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,balanced,0.08470400174458821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,balanced,0.09293333689371745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,balanced,0.1083519955476125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.113946666320165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.1459999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.16275733709335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.20965866247812906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.2607626716295878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.36735467116038006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.46876267592112225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.66811203956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.8732213179270426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.08316159844398499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.07988479733467102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.08958719968795777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.09502720236778259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.14307839870452882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.15319679975509642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.19081599712371827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.21451520919799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.28395519256591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.34410240650177004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.482860803604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,power_law_1.01,0.6368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,power_law_1.01,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,power_law_1.01,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,power_law_1.01,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,power_law_1.01,0.02964479923248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,power_law_1.01,0.02995840013027191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,power_law_1.01,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,power_law_1.01,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,power_law_1.01,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,power_law_1.01,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,power_law_1.01,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,power_law_1.01,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,power_law_1.01,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,power_law_1.01,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,power_law_1.01,0.1365823984146118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,power_law_1.01,0.17783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,power_law_1.01,0.20801279544830323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,power_law_1.01,0.30693759918212893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,power_law_1.01,0.3894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,power_law_1.2,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,power_law_1.2,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,power_law_1.2,0.05764480233192444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,power_law_1.2,0.0580735981464386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,power_law_1.2,0.06398079991340637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,power_law_1.2,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,power_law_1.2,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,power_law_1.2,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,power_law_1.2,0.09770240187644959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,power_law_1.2,0.11474560499191284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,power_law_1.2,0.13411840200424194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,power_law_1.2,0.16410239934921264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,power_law_1.2,0.21822080612182618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,power_law_1.2,0.25849599838256837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,power_law_1.2,0.4266240119934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,power_law_1.2,0.5143167972564697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,power_law_1.2,0.6870528221130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,power_law_1.2,0.9341119766235352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,balanced,0.05144000053405762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,balanced,0.06853866577148438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,balanced,0.07549333572387695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,balanced,0.08919466535250346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,balanced,0.10331199566523235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,balanced,0.13329066832860312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,balanced,0.1649066706498464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,balanced,0.22359466552734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,balanced,0.28302399317423504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,balanced,0.04473066826661428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,balanced,0.04533866544564565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,balanced,0.05418133238951365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,balanced,0.0574186642964681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,balanced,0.0768746683994929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,balanced,0.07664533456166585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,balanced,0.08627733588218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,balanced,0.09077333410580952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,balanced,0.1072213351726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,balanced,0.12127466996510823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,balanced,0.1564853290716807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,balanced,0.17711466550827026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,balanced,0.22643733024597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,balanced,0.271397332350413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,balanced,0.36737600962320965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,balanced,0.4627306858698527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,power_law_1.2,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,power_law_1.2,0.14455679655075074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,power_law_1.2,0.1562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,power_law_1.2,0.2218240022659302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,power_law_1.2,0.234771203994751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,power_law_1.2,0.2977344036102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,power_law_1.2,0.3915071964263916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,power_law_1.2,0.34445440769195557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,power_law_1.2,0.37343358993530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,power_law_1.2,0.3657344102859497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,power_law_1.2,0.398854398727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,power_law_1.2,0.3992255926132202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,power_law_1.2,0.4028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,power_law_1.2,0.42115201950073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,power_law_1.2,0.4388095855712891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,power_law_1.2,0.4574272155761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,power_law_1.2,0.48024959564208985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,power_law_1.2,0.5160255908966065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,power_law_1.2,0.5226624011993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,power_law_1.2,0.6436736106872558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,power_law_1.2,0.6483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,power_law_1.2,0.797983980178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,power_law_1.2,0.884217643737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,power_law_1.2,1.1102272033691407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,power_law_1.2,1.3354816436767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,power_law_1.2,1.736332893371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,power_law_1.2,2.304652786254883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,balanced,0.02532266577084859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,balanced,0.027269333600997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,balanced,0.025226667523384094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,balanced,0.027322667340437572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,balanced,0.03303466737270355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,balanced,0.03604800005753835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,balanced,0.049285332361857094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,balanced,0.06914133330186208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,balanced,0.1018506685892741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,balanced,0.12586667140324911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,balanced,0.14852266510327658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,balanced,0.037978666524092354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,balanced,0.03793599953254064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,balanced,0.03806400050719579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,balanced,0.04345599810282389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,balanced,0.04348800083001455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,balanced,0.049125333627065025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,balanced,0.05625066657861074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,balanced,0.0662666658560435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,balanced,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,balanced,0.11619733770688374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,balanced,0.14816000064214072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,balanced,0.17683732509613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.1046336054801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.10139520168304443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.11935360431671142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.12590080499649048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.12469120025634765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.1223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.12591359615325928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.13744640350341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.1435520052909851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.1565119981765747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.17407360076904296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.19503359794616698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.22751359939575194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.26229760646820066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.32857599258422854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.3911936044692993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.5397823810577392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.6623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.9574144363403321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.2861696243286134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.0939520001411438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.11070719957351685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.1383936047554016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.20930559635162355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.24518399238586425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.26617600917816164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.26296319961547854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.26299519538879396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.27562880516052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.279584002494812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.2879807949066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.30766079425811765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.3147072076797485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.3230079889297485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.34749441146850585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.3770431995391846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.42287359237670896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.48233599662780763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.5354944229125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.6765183925628662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.7697792053222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.0107263565063476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.2596799850463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.6921600341796874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.2911808013916017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,power_law_1.01,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,power_law_1.01,0.06378239989280701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,power_law_1.01,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,power_law_1.01,0.09546239972114563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,power_law_1.01,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,power_law_1.01,0.14824960231781006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,power_law_1.01,0.2006848096847534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,power_law_1.01,0.2542912006378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,power_law_1.01,0.3764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,power_law_1.01,0.4686016082763672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,balanced,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,balanced,0.032586666444937386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,balanced,0.03323200096686681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,balanced,0.0407679999868075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,balanced,0.08089600006739299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,balanced,0.0909493366877238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,balanced,0.12173866232236226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,balanced,0.13849600156148276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,balanced,0.18226132790247598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,balanced,0.2179093360900879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,power_law_1.01,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,power_law_1.01,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,power_law_1.01,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,power_law_1.01,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,power_law_1.01,0.09202560186386108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,power_law_1.01,0.0951744019985199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,power_law_1.01,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,power_law_1.01,0.10340479612350464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,power_law_1.01,0.1115455985069275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,power_law_1.01,0.12267520427703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,power_law_1.01,0.13810559511184692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,power_law_1.01,0.15533440113067626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,power_law_1.01,0.19028480052948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,power_law_1.01,0.21868159770965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,power_law_1.01,0.30205440521240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,power_law_1.01,0.3715008020401001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,power_law_1.01,0.4994688034057617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,power_law_1.01,0.6655104160308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,power_law_1.01,0.8914239883422852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,power_law_1.01,1.1850111961364747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.027539199590682982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.02908160090446472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.02906239926815033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.029491201043128967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.02956160008907318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.05248640179634094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.10122879743576049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.11589759588241577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.15624959468841554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,balanced,0.07437333464622498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,balanced,0.07630399862925212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,balanced,0.09063466389973958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,balanced,0.12995200355847678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,balanced,0.1909653345743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,balanced,0.2487199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,balanced,0.2419253389040629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,balanced,0.24278400341669717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,balanced,0.24280534187952676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,balanced,0.2482773264249166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,balanced,0.25015467405319214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,balanced,0.2495680054028829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,balanced,0.25492266813913983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,balanced,0.26473067204157513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,balanced,0.26424533128738403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,balanced,0.2759573260943095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,balanced,0.2990399996439616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.31617599725723267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.36212265491485596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.3978186845779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.48951999346415204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.5801440080006918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.7696853478749593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,balanced,1.0248640378316243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,balanced,1.37882661819458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,balanced,1.8515572547912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.2620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,power_law_1.01,0.3345855951309204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,balanced,0.06524799764156342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,balanced,0.09690133730570476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,balanced,0.0990559955437978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,balanced,0.10166399677594502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,balanced,0.09943999846776326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,balanced,0.10106666882832845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,balanced,0.10139733552932739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,balanced,0.1034399966398875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,balanced,0.1035040020942688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,balanced,0.1074666678905487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,balanced,0.10928533474604289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,balanced,0.11122666796048482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,balanced,0.11909866333007812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,balanced,0.12541332840919495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,balanced,0.13923733433087668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,balanced,0.15190399686495462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,balanced,0.17997332413991293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,balanced,0.20371200640996298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,balanced,0.2733599940935771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,balanced,0.3325226704279582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,balanced,0.48154131571451825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,balanced,0.6219840049743652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.14117759466171265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.1429952025413513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.14353920221328736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.1333184003829956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.09315199851989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.10748800039291381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.10972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.1263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.12059520483016968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.1278272032737732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.12266240119934083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.1266816020011902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.14067840576171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.1299008011817932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.14056320190429689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.13267840147018434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.13470720052719115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.13274879455566407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.14024319648742675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.1373952031135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.14324480295181274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.13841919898986815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.15146880149841307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.15352319478988646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.1555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.1689792037010193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.18150399923324584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.19648000001907348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.17930879592895507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.23454079627990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.15999360084533693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.24292480945587158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.18053120374679565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,power_law_1.2,0.09674879908561707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.31007359027862547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.1893183946609497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,power_law_1.2,0.11978240013122558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.3739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,power_law_1.2,0.16252800226211547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.46787199974060056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.6202432155609131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.8611392021179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.22234880924224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.1113792419433595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.2344127893447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,power_law_1.2,0.2048703908920288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.2729151964187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,power_law_1.2,0.2510783910751343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.3173696041107178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.4137599945068359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,power_law_1.2,0.3517119884490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,power_law_1.01,0.5071231842041015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,power_law_1.2,0.36588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,power_law_1.01,0.716428804397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,power_law_1.2,0.3494271993637085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,power_law_1.2,0.36445438861846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,power_law_1.01,0.8441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,power_law_1.2,0.38638079166412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,power_law_1.2,0.40563201904296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,power_law_1.2,0.4209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,power_law_1.2,0.4244095802307129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,power_law_1.2,0.4521920204162598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,power_law_1.2,0.46129279136657714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,power_law_1.2,0.47813758850097654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,power_law_1.2,0.5334335803985596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,power_law_1.2,0.575648021697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,power_law_1.2,0.6620800018310546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,power_law_1.2,0.6900159835815429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,power_law_1.2,0.8400447845458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,power_law_1.2,0.9524288177490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,power_law_1.2,1.2459199905395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,power_law_1.2,1.5019136428833009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,power_law_1.2,2.131648063659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,power_law_1.2,2.659872055053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,power_law_1.2,0.15215359926223754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,power_law_1.2,0.2536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,power_law_1.2,0.36907520294189455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,power_law_1.2,0.510265588760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,power_law_1.2,0.8808768272399903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,power_law_1.2,0.913548755645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,power_law_1.2,0.9675904273986816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,power_law_1.2,0.9779711723327636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,power_law_1.2,1.0202943801879882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,power_law_1.2,1.0597503662109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,power_law_1.2,1.044281578063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,power_law_1.2,1.0786239624023437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,power_law_1.2,1.1127552032470702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,power_law_1.2,1.1557696342468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,power_law_1.2,1.1661120414733888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,power_law_1.2,1.203264045715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,power_law_1.2,1.3185152053833007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,power_law_1.2,1.402444839477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,power_law_1.2,1.6304256439208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,power_law_1.2,1.762175941467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,power_law_1.2,2.153875160217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,power_law_1.2,2.2524864196777346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,power_law_1.2,2.770124816894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,power_law_1.2,3.682233428955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,power_law_1.2,4.726054382324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,power_law_1.2,5.705171203613281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,power_law_1.2,0.11038080453872681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,power_law_1.2,0.10575360059738159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,power_law_1.2,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,power_law_1.2,0.10791679620742797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,power_law_1.2,0.10259840488433838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,power_law_1.2,0.09056000113487243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,power_law_1.2,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.01,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.01,0.02930560111999512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,power_law_1.2,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.01,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,power_law_1.2,0.11104639768600463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.01,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,power_law_1.2,0.1056831955909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.01,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,power_law_1.2,0.11226880550384521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.01,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,power_law_1.2,0.11004799604415894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.01,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,power_law_1.2,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.01,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.01,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,power_law_1.2,0.11822079420089722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,power_law_1.2,0.12199679613113404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.01,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,power_law_1.2,0.1210752010345459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.01,0.06862720251083373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,power_law_1.2,0.1293887972831726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.01,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,power_law_1.2,0.14469759464263915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.01,0.10061440467834473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,power_law_1.2,0.15626879930496215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.01,0.12539520263671874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.01,0.1464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,power_law_1.2,0.1889024019241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.01,0.16100480556488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,power_law_1.2,0.22617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,power_law_1.2,0.28977279663085936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,power_law_1.2,0.339737606048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,power_law_1.2,0.4807231903076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,power_law_1.2,0.5757184028625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,power_law_1.2,0.842080020904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,power_law_1.2,1.1085439682006837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,balanced,0.0787360022465388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,balanced,0.08143466711044312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,balanced,0.08099733293056488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,balanced,0.08177066842714946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,balanced,0.08262933293978374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,balanced,0.08572266499201457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,balanced,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,balanced,0.08727999528249104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,balanced,0.09012800455093384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,balanced,0.09680533409118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,balanced,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,balanced,0.10717866818110149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,balanced,0.1186186671257019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,balanced,0.12787200013796488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,balanced,0.14908799529075623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,balanced,0.182751993338267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,balanced,0.22285334269205728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,balanced,0.26526933908462524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,balanced,0.3587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,balanced,0.4553813139597575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,balanced,0.6401919921239217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,balanced,0.8256853421529134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,balanced,0.05771199862162272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,balanced,0.058970664938290916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,balanced,0.05942933261394501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,balanced,0.06113600234190623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,balanced,0.06145066519578298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,balanced,0.06261866788069408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,balanced,0.06605866551399231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,balanced,0.06804266571998596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,balanced,0.08090666433175404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,balanced,0.0802400012811025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,balanced,0.0846720039844513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,balanced,0.10300266742706299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,balanced,0.10652800401051839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,balanced,0.13384000460306802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,balanced,0.16260266304016113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,balanced,0.213536004225413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,balanced,0.2515786687533061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,balanced,0.3562026818593343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,balanced,0.45181334018707275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,balanced,0.6465333302815756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,balanced,0.8426773548126221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.08045439720153809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.1207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.1656448006629944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.2032831907272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.26784000396728513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.35199999809265137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,balanced,0.1448746621608734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,balanced,0.24625066916147867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,balanced,0.24612800280253092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,balanced,0.24480533599853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,balanced,0.24728000164031982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,balanced,0.24799466133117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,balanced,0.25013333559036255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,balanced,0.25225599606831867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,balanced,0.2525493303934733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,balanced,0.25650666157404584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,balanced,0.2634399930636088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,balanced,0.2665440042813619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,balanced,0.27327466011047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,balanced,0.28732266028722125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,balanced,0.29867200056711835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,balanced,0.3218986590703328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,balanced,0.34867199261983234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,balanced,0.3948586781819661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,balanced,0.4374399979909261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,balanced,0.565445343653361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,balanced,0.647050658861796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,balanced,0.8975040117899576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.14584319591522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,balanced,1.0653013388315837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.24167680740356445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.29576959609985354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.4790016174316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.6002048015594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.7151487827301025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,power_law_1.01,0.9017919540405274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,power_law_1.01,0.8706496238708497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,power_law_1.01,0.9134464263916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,power_law_1.01,0.9355903625488281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,power_law_1.01,0.9889856338500976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,power_law_1.01,0.9873920440673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,power_law_1.01,1.0219136238098145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,power_law_1.01,1.0218751907348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.054297637939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.0793984413146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.1948415756225585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.2319104194641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.390732765197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,power_law_1.01,1.4017279624938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,power_law_1.01,1.5934975624084473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.4309696197509766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,power_law_1.01,1.869913673400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,power_law_1.01,2.2961984634399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,power_law_1.01,2.340422439575195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,power_law_1.01,2.8887296676635743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,power_law_1.01,3.6663936614990233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.5550464153289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,power_law_1.2,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,power_law_1.2,0.10840320587158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,power_law_1.2,0.15449600219726561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,power_law_1.2,0.22145919799804686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,power_law_1.2,0.2848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,power_law_1.2,0.33792641162872317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,power_law_1.2,0.45863680839538573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,power_law_1.2,0.4740096092224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,power_law_1.2,0.47762560844421387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,power_law_1.2,0.5128384113311768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,power_law_1.2,0.537113618850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,power_law_1.2,0.5525440216064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,power_law_1.2,0.569542407989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,power_law_1.2,0.5941120147705078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,power_law_1.2,0.6068096160888672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,power_law_1.2,0.646015977859497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,power_law_1.2,0.7058752059936524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,power_law_1.2,0.783897590637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,power_law_1.2,0.928927993774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,power_law_1.2,1.0114175796508789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,power_law_1.2,1.2414400100708007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,power_law_1.2,1.4798591613769532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,power_law_1.2,1.9059520721435548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,power_law_1.2,2.2955839157104494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,power_law_1.2,3.025996780395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,power_law_1.2,3.947609710693359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,power_law_1.01,0.07078400254249573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,power_law_1.01,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,power_law_1.01,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,power_law_1.01,0.073580801486969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,power_law_1.01,0.07459200024604798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,power_law_1.01,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,power_law_1.01,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,power_law_1.01,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,power_law_1.01,0.08550400137901307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,power_law_1.01,0.08836479783058167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,power_law_1.01,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,power_law_1.01,0.09820799827575684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,power_law_1.01,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,power_law_1.01,0.10432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,power_law_1.01,0.12031999826431275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.01,0.141593599319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.01,0.17637759447097778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.01,0.20543360710144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.01,0.29044480323791505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.01,0.3288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.01,0.45815038681030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.01,0.5836095809936523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.01,0.8724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.01,1.1595392227172852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,balanced,0.03286399940649668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,balanced,0.05787200232346853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,balanced,0.07973866661389668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,balanced,0.08470400174458821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,balanced,0.08455999692281087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,balanced,0.08569066723187764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,balanced,0.0841439962387085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,balanced,0.0879146655400594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,balanced,0.0904960036277771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,balanced,0.09017067154248555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,balanced,0.09773866335550944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,balanced,0.09909866253534953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,balanced,0.10505599776903789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,balanced,0.1130400002002716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,balanced,0.12247999509175618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,balanced,0.14495467146237692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,balanced,0.15833066900571188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,balanced,0.2041119933128357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,balanced,0.23731732368469238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,balanced,0.33345599969228107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,balanced,0.4092213312784831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,balanced,0.5764266649881998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,balanced,0.7428533236185709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,power_law_1.01,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,power_law_1.01,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,power_law_1.01,0.12077440023422241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,power_law_1.01,0.1183359980583191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,power_law_1.01,0.11550719738006592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,power_law_1.01,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,power_law_1.01,0.11742719411849975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,power_law_1.01,0.11415679454803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,power_law_1.01,0.11153919696807861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,power_law_1.01,0.11817599534988403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,power_law_1.01,0.12199679613113404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,power_law_1.01,0.12293119430541992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,power_law_1.01,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,power_law_1.01,0.13248000144958497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,power_law_1.01,0.14085760116577148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,power_law_1.01,0.1661631941795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,power_law_1.01,0.17511039972305298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,power_law_1.01,0.20034561157226563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,power_law_1.01,0.2347775936126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,power_law_1.01,0.2966399908065796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,power_law_1.01,0.34824318885803224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,power_law_1.01,0.4627903938293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,power_law_1.01,0.5664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,power_law_1.01,0.7831935882568359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,power_law_1.01,1.0522687911987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.07417600154876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.08531200289726257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.08209279775619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.08936960101127625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.15184639692306517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.18672640323638917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.20799999237060546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.27765119075775146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.3442431926727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,power_law_1.01,0.45494399070739744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,power_law_1.01,0.6152959823608398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.01,0.09348480105400085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.01,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.01,0.12236160039901733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.01,0.1518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.01,0.14766080379486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.01,0.1449280023574829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.01,0.14590079784393312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.01,0.15476479530334472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.01,0.1522495985031128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.01,0.15898239612579346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.01,0.16211199760437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.01,0.1681663990020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.01,0.17860480546951293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.01,0.18094719648361207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.01,0.19105279445648193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.01,0.21960959434509278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.01,0.24154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.01,0.2957632064819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.01,0.36357760429382324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.01,0.46978559494018557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.01,0.5286399841308593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.01,0.7024511814117431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.01,0.8718144416809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.01,1.216268825531006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.01,1.5662464141845702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,balanced,0.07391466697057088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,balanced,0.10310932993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,balanced,0.154448002576828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,balanced,0.25409066677093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,balanced,0.44747201601664227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,balanced,0.4538453420003255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,balanced,0.45527998606363934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,balanced,0.4607146581013997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,balanced,0.4642080068588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,balanced,0.4656960169474284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,balanced,0.4800106684366862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,balanced,0.48323198159535724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,balanced,0.49026668071746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,balanced,0.5035733381907145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,balanced,0.5073333183924357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,balanced,0.5256266593933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,balanced,0.5608213345209757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,balanced,0.5902133385340372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,balanced,0.6627519925435384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,balanced,0.7351840337117513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,balanced,0.9302133719126383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,balanced,1.0328266620635986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,balanced,1.4144213994344075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,balanced,1.6984052658081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,balanced,2.3891146977742515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,balanced,3.0060907999674478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,power_law_1.01,0.06837760210037232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,power_law_1.01,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,power_law_1.01,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,power_law_1.01,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,power_law_1.01,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,power_law_1.01,0.10118399858474732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,power_law_1.01,0.11133439540863037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,power_law_1.01,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,power_law_1.01,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,power_law_1.01,0.20157439708709718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,power_law_1.01,0.25265278816223147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,power_law_1.01,0.3410624027252197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,power_law_1.01,0.43549442291259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,power_law_1.01,0.6429183959960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,power_law_1.01,0.847475242614746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,balanced,0.054474666714668274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,balanced,0.05568000177542368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,balanced,0.06295999884605408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,balanced,0.08513066172599792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,balanced,0.11799466609954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,balanced,0.1762453317642212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,balanced,0.2381440003712972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,balanced,0.23757867018381754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,balanced,0.2369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,balanced,0.23611734310785928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,balanced,0.2403200070063273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,balanced,0.24165334304173788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,balanced,0.2442880074183146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,balanced,0.2453119953473409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,balanced,0.25381867090861004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,balanced,0.25618666410446167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,balanced,0.26129066944122314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,balanced,0.2718399961789449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.28200000524520874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.3028480013211568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.3254879911740621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.3715519905090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.41701332728068036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.5100586811701456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.6161760091781616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,balanced,0.7897546291351318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,balanced,1.03657062848409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,power_law_1.2,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,power_law_1.2,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,power_law_1.2,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,power_law_1.2,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,power_law_1.2,0.05915520191192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,power_law_1.2,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,power_law_1.2,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,power_law_1.2,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,power_law_1.2,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,power_law_1.2,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,power_law_1.2,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,power_law_1.2,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,power_law_1.2,0.16670080423355102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,power_law_1.2,0.21444480419158934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,power_law_1.2,0.22421760559082032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,power_law_1.2,0.23119359016418456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,power_law_1.2,0.23632640838623048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,power_law_1.2,0.24296960830688477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,power_law_1.2,0.25350399017333985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,power_law_1.2,0.25772800445556643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,power_law_1.2,0.2684607982635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,power_law_1.2,0.2802047967910767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,power_law_1.2,0.2844032049179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,power_law_1.2,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,power_law_1.2,0.30117120742797854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,power_law_1.2,0.332096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.2,0.36306560039520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.2,0.4281023979187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,power_law_1.2,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.2,0.5015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.2,0.6311359882354737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,power_law_1.2,0.16695040464401245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.2,0.7563136100769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.2,1.0306816101074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.2,1.278335952758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.2,1.7925952911376952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.2,2.3053119659423826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,power_law_1.2,0.21909759044647217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,power_law_1.2,0.26204159259796145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,power_law_1.2,0.36420478820800783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,power_law_1.2,0.4582655906677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,power_law_1.2,0.6560895919799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,balanced,0.058464000622431435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,balanced,0.07460799813270569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,balanced,0.07790933549404144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,balanced,0.07916800181070964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,balanced,0.07844799757003784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,balanced,0.08363200227419536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,balanced,0.08488000432650249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,balanced,0.09016533692677815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,balanced,0.09728533029556274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,balanced,0.10076799988746643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,balanced,0.1092800001303355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,balanced,0.1269493301709493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,balanced,0.13804266850153604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,balanced,0.16664533813794455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,balanced,0.19028266270955405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,balanced,0.26600533723831177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,balanced,0.31817599137624103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,balanced,0.45477867126464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,balanced,0.5777279933293661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,balanced,0.8367466926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,balanced,1.0911040306091309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.09696000218391418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.1422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.16142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.20182399749755858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.26864640712738036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.3112191915512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.37460479736328123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.4060031890869141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.3485759973526001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.3504767894744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.3859071969985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.3939968109130859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.3925120115280151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.3965503931045532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.4031424045562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.44759039878845214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.4618112087249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.5015039920806885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.5241600036621094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.5868800163269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,power_law_1.2,0.5618112087249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,power_law_1.2,0.7253568172454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,power_law_1.2,0.8122048377990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,power_law_1.2,1.0652159690856933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,power_law_1.2,1.154860782623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,power_law_1.2,1.6004287719726562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,power_law_1.2,1.9106239318847655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,power_law_1.2,0.8477696418762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,power_law_1.01,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,power_law_1.01,0.04433279931545257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,power_law_1.01,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,power_law_1.01,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,power_law_1.01,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,power_law_1.01,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,power_law_1.01,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,power_law_1.01,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,power_law_1.01,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,power_law_1.01,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,power_law_1.01,0.1002303957939148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,power_law_1.01,0.12264319658279418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,power_law_1.01,0.13971840143203734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,power_law_1.01,0.18394880294799804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,power_law_1.01,0.21120638847351075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,power_law_1.01,0.29745280742645264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,power_law_1.01,0.4100351810455322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,power_law_1.01,0.5611328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,power_law_1.01,0.6640639781951905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,power_law_1.01,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,power_law_1.01,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,power_law_1.01,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,power_law_1.01,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,power_law_1.01,0.07100160121917724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,power_law_1.01,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,power_law_1.01,0.08743680119514466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,power_law_1.01,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,power_law_1.01,0.13241599798202514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,power_law_1.01,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,power_law_1.01,0.19237760305404664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,power_law_1.01,0.2452672004699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,power_law_1.01,0.34729599952697754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,power_law_1.01,0.4162112236022949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,power_law_1.01,0.6198272228240966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,power_law_1.01,0.7870719909667969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.18335360288619995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.3162175893783569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.4751743793487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.2,0.6443264007568359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.200864028930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.2,1.2910016059875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.2,1.3546815872192384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.2,1.405395221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.2,1.4376832008361817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.2,1.4863743782043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.2,1.512838363647461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.2,1.5614527702331542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.2,1.593228816986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.2,1.659449577331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.2,1.7049856185913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.2,1.7925823211669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.2,1.9410367965698243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.2,2.002412796020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.061843198537826535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.2,2.3023296356201173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.06798080205917359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.2,2.5020288467407226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.07705600261688232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.11383680105209351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.2,3.0518976211547852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.14064639806747437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.16872960329055786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.2310784101486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.29419519901275637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.2,3.076576042175293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.3984639883041382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.5186048030853272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.7879168033599854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.2,3.7379966735839845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,0.9747391700744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.2,4.752127838134766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.2,6.573811340332031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.2,7.495263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,power_law_1.01,0.022867199778556824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,power_law_1.01,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,power_law_1.01,0.023712000250816344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,power_law_1.01,0.02401279956102371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,power_law_1.01,0.0247871994972229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,power_law_1.01,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,power_law_1.01,0.025068798661231996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,power_law_1.01,0.026675200462341307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,power_law_1.01,0.028204798698425293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,power_law_1.01,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,power_law_1.01,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.14736640453338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.1679808020591736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,power_law_1.2,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,power_law_1.2,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,power_law_1.2,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,power_law_1.2,0.08906239867210389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,power_law_1.2,0.08791040182113648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,power_law_1.2,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,power_law_1.2,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,power_law_1.2,0.09048960208892823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,power_law_1.2,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,power_law_1.2,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,power_law_1.2,0.10219520330429077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,power_law_1.2,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,power_law_1.2,0.11432960033416747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.2,0.12490880489349365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.2,0.15291520357131957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.2,0.17327359914779664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.2,0.22254080772399903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.2,0.2624063968658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.2,0.3563647985458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.2,0.4953472137451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.2,0.7472896099090576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.2,0.9889727592468261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,balanced,0.05634133517742157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,balanced,0.0722453345855077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,balanced,0.07471466561158498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,balanced,0.0765119989713033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,balanced,0.07855999966462453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,balanced,0.08611200253168742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,balanced,0.08917333682378133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,balanced,0.09900266925493877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,balanced,0.10686399539311726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,balanced,0.12454932928085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,balanced,0.15074132879575095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,balanced,0.18398400147755942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,balanced,0.21174933513005575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,balanced,0.29335467020670575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,balanced,0.36070934931437176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,balanced,0.5177280108133951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,balanced,0.6634399890899658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,power_law_1.01,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,power_law_1.01,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,power_law_1.01,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,power_law_1.01,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,power_law_1.01,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,power_law_1.01,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,power_law_1.01,0.07775359749794006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,power_law_1.01,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,power_law_1.01,0.12522879838943482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,power_law_1.01,0.17994240522384644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,power_law_1.01,0.216377592086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,power_law_1.01,0.3254271984100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,power_law_1.01,0.46467838287353513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,balanced,0.04614399870236715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,balanced,0.028517333169778187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,balanced,0.029701332251230877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,balanced,0.031152000029881794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,balanced,0.041178666055202484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,balanced,0.041450666884581246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,balanced,0.06414400041103363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,balanced,0.03443733354409536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,balanced,0.07072000205516815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,balanced,0.07797333101431529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,balanced,0.07997333506743114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,balanced,0.08015466729799907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,balanced,0.08110400040944417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,balanced,0.08109866579373677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,balanced,0.08081600069999695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,balanced,0.0825973351796468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,balanced,0.08275733391443889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,balanced,0.08593066533406575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,balanced,0.08708799878756206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,balanced,0.08955732981363933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,balanced,0.10575466354688008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,balanced,0.10194133718808492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,balanced,0.13595733046531677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,balanced,0.1285866697629293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,balanced,0.1694613297780355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,balanced,0.18516266345977783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,balanced,0.25142399470011395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,balanced,0.2882879972457886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,balanced,0.4087626536687215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,balanced,0.5015199979146322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,balanced,0.07844799757003784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,balanced,0.05364799996217092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,balanced,0.05884799857934316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,balanced,0.06502933303515117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,balanced,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,balanced,0.08425600330034892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,balanced,0.08386666576067607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,balanced,0.10727999607721965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,balanced,0.11966933806737264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,balanced,0.15661866466204324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,balanced,0.14477333426475525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,balanced,0.18480000893274942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,balanced,0.19562667608261108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,balanced,0.2622080047925313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,balanced,0.2985493342081706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,balanced,0.05340266724427541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,balanced,0.08974933624267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,balanced,0.08782399694124858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,balanced,0.09084266424179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,balanced,0.09113066395123799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,balanced,0.08910933136940002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,balanced,0.09231467048327129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,balanced,0.09275733431180318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,balanced,0.09522666533788045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,balanced,0.09589866797129314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,balanced,0.09899733463923137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,balanced,0.10272533694903056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,balanced,0.10641599694887798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,balanced,0.10944533348083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,balanced,0.11183999975522359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,balanced,0.14046933253606161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,balanced,0.15029333035151163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,balanced,0.19326933224995932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,balanced,0.24370666344960532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,balanced,0.3471999963124593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,balanced,0.4568800131479899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,balanced,0.6553813219070435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,balanced,0.8591307004292806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,balanced,1.227952003479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,balanced,1.602735996246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,power_law_1.01,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,power_law_1.01,0.1428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,power_law_1.01,0.18403199911117554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,power_law_1.01,0.2252415895462036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,power_law_1.01,0.2648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,power_law_1.01,0.3092223882675171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,power_law_1.01,0.3710464000701904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,power_law_1.01,0.3724735975265503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,power_law_1.01,0.36348800659179686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,power_law_1.01,0.37800960540771483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,power_law_1.01,0.39423360824584963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,power_law_1.01,0.39431679248809814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,power_law_1.01,0.3903039932250977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,power_law_1.01,0.40778241157531736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,power_law_1.01,0.4324031829833984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,power_law_1.01,0.44063358306884765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,power_law_1.01,0.4574592113494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,power_law_1.01,0.5092671871185303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,power_law_1.01,0.5007552146911621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,power_law_1.01,0.5899136066436768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,power_law_1.01,0.6085696220397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,power_law_1.01,0.7287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,power_law_1.01,0.8401856422424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,power_law_1.01,1.0768511772155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,power_law_1.01,1.2264960289001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,power_law_1.01,1.701273536682129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,power_law_1.01,2.0843904495239256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.08307200074195861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.10990079641342163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.12473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.13575040102005004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.16471680402755737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.16722559928894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.156550395488739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.17427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.16666239500045776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.1692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.17726080417633056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.18599679470062255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.193996798992157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.20336639881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.21417601108551027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.24270079135894776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.26423680782318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.32198400497436525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.3338943958282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.43564162254333494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.528108787536621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.673363208770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.8320575714111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.1835328102111817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.438169574737549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,power_law_1.2,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,power_law_1.2,0.028896000981330872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,power_law_1.2,0.03018240034580231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,power_law_1.2,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,power_law_1.2,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,power_law_1.2,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,power_law_1.2,0.15121279954910277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,power_law_1.2,0.17953920364379883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,power_law_1.2,0.24172799587249755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,power_law_1.01,0.14817919731140136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,power_law_1.01,0.24573440551757814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,power_law_1.01,0.21062400341033935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,power_law_1.2,0.3282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,power_law_1.01,0.26898560523986814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,power_law_1.2,0.058931201696395874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,power_law_1.01,0.3244352102279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,power_law_1.2,0.10436480045318604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,power_law_1.01,0.31025280952453616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,power_law_1.2,0.13523199558258056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,power_law_1.01,0.3321791887283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,power_law_1.2,0.16885119676589966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,power_law_1.01,0.28886399269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,power_law_1.2,0.2257535934448242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,power_law_1.01,0.2820800065994263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,power_law_1.2,0.2310528039932251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,power_law_1.2,0.24167039394378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,power_law_1.01,0.3007296085357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,power_law_1.2,0.24977281093597412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,power_law_1.01,0.3097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,power_law_1.2,0.25823359489440917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,power_law_1.01,0.3015552043914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,power_law_1.2,0.2682687997817993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,power_law_1.01,0.2986687898635864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,power_law_1.2,0.28166399002075193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,power_law_1.01,0.326310396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,power_law_1.01,0.3085439920425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,power_law_1.01,0.34158079624176024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,power_law_1.01,0.36430718898773196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,power_law_1.01,0.3953727960586548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,power_law_1.2,0.30207359790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,power_law_1.01,0.4701056003570557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,power_law_1.2,0.32503681182861327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,power_law_1.01,0.469209623336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,power_law_1.2,0.33632640838623046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,power_law_1.2,0.3487423896789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,power_law_1.01,0.589676809310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,power_law_1.2,0.40872960090637206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,power_law_1.01,0.6544447898864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,power_law_1.2,0.46230401992797854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,power_law_1.01,0.8497856140136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,power_law_1.2,0.5624639987945557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,power_law_1.01,0.9589695930480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,power_law_1.2,0.6233856201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,power_law_1.01,1.2534144401550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,power_law_1.2,0.7968575954437256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,power_law_1.01,1.6241920471191407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,power_law_1.2,0.9782784461975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,power_law_1.2,1.3169983863830566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,power_law_1.2,1.6523263931274415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,power_law_1.2,2.303436851501465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,power_law_1.2,2.9863679885864256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.12386560440063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.15203200578689574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.16672639846801757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.1775488018989563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.21519360542297364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.2066943883895874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.2197376012802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.20551040172576904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.20263679027557374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.2144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.23277440071105956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.23982079029083253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.2518784046173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.26190719604492185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.2816512107849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.31427199840545655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.3747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.4191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.5408703804016113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.5810304164886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.758406400680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,0.9264575958251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.2957311630249024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,1.6557056427001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,power_law_1.01,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,power_law_1.01,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,power_law_1.01,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,power_law_1.01,0.09599360227584838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,power_law_1.01,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,power_law_1.01,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,power_law_1.01,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,power_law_1.01,0.11817599534988403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,power_law_1.01,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,power_law_1.01,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,power_law_1.01,0.12810239791870118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,power_law_1.01,0.13889919519424437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,power_law_1.01,0.17872639894485473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.01,0.20769920349121093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.01,0.26832640171051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.01,0.35507841110229493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.01,0.41278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.01,0.5322624206542969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.01,0.8217344284057617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.01,1.112326431274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.01,1.4799360275268554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.01,2.2840703964233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,balanced,0.08730666836102803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,balanced,0.1099626620610555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,balanced,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,balanced,0.24465600649515787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,balanced,0.4076319932937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,balanced,0.41282133261362713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,balanced,0.4140586853027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,balanced,0.4136586586634318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,balanced,0.4166826804478963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,balanced,0.4235359827677409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,balanced,0.4265386660893758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,balanced,0.43029332160949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,balanced,0.43587732315063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,balanced,0.4443519910176595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,balanced,0.4500266710917155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,balanced,0.46306665738423664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,balanced,0.49774932861328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,balanced,0.5252373218536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,balanced,0.5994879802068075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,balanced,0.6652426719665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,balanced,0.7808693250020345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,balanced,0.9043839772542318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,balanced,1.2649386723836262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,balanced,1.4797600110371907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,balanced,2.137061278025309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,balanced,2.6862878799438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,balanced,0.03852266569932302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,balanced,0.055311997731526695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,balanced,0.07763200004895528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,balanced,0.12298132975896199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,balanced,0.20560532808303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,balanced,0.210533340771993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,balanced,0.21238933006922403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,balanced,0.21322667598724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,balanced,0.21203200022379556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,balanced,0.22155199448267618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,balanced,0.21555733680725098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,balanced,0.22294400135676065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,balanced,0.2271733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,balanced,0.22932799657185873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,balanced,0.23771200577418009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,balanced,0.24363734324773154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,balanced,0.25462400913238525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,balanced,0.2727573315302531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,balanced,0.29529066880544025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,balanced,0.3619360129038493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,balanced,0.3959840138753255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,balanced,0.5168373187383016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,balanced,0.5900693337122599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,balanced,0.816981315612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,balanced,0.9913120269775391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,balanced,1.4293707211812336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,balanced,1.8151626586914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,power_law_1.01,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,power_law_1.01,0.12842880487442015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,power_law_1.01,0.11928319931030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,power_law_1.01,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,power_law_1.01,0.11962239742279053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,power_law_1.01,0.11113599538803101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,power_law_1.01,0.11362559795379638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,power_law_1.01,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,power_law_1.01,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,power_law_1.01,0.1087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,power_law_1.01,0.1103935956954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,power_law_1.01,0.11361279487609863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,power_law_1.01,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,power_law_1.01,0.13621759414672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,power_law_1.01,0.13443200588226317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,power_law_1.01,0.14599039554595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,power_law_1.01,0.15854079723358155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,power_law_1.01,0.17056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,power_law_1.01,0.21449599266052247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,power_law_1.01,0.22836480140686036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,power_law_1.01,0.29949440956115725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,power_law_1.01,0.32636160850524903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,power_law_1.01,0.44179201126098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,power_law_1.01,0.5445631980895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,power_law_1.01,0.7724736213684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,power_law_1.01,0.935148811340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,power_law_1.2,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,power_law_1.2,0.16183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,power_law_1.2,0.20000638961791992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,power_law_1.2,0.3784064054489136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,power_law_1.2,0.3727423906326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,power_law_1.2,0.3768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,power_law_1.2,0.393337607383728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,power_law_1.2,0.39368960857391355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,power_law_1.2,0.39782400131225587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,power_law_1.2,0.4074751853942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,power_law_1.2,0.4249152183532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,power_law_1.2,0.44503040313720704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,power_law_1.2,0.4411776065826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,power_law_1.2,0.44275197982788084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,power_law_1.2,0.48107519149780276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,power_law_1.2,0.4866623878479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,power_law_1.2,0.5226175785064697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,power_law_1.2,0.5907008171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,power_law_1.2,0.6913472175598144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,power_law_1.2,0.8448639869689941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,power_law_1.2,0.9751104354858399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,power_law_1.2,1.296940803527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,power_law_1.2,1.6351615905761718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,power_law_1.2,2.3207231521606446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,power_law_1.2,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,power_law_1.2,0.03415679931640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,power_law_1.2,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,power_law_1.2,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,power_law_1.2,3.007609558105469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,power_law_1.2,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,power_law_1.2,0.12445440292358398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,power_law_1.2,0.18284800052642822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,power_law_1.2,4.329497528076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,power_law_1.2,0.21218559741973878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,power_law_1.2,0.27991039752960206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,power_law_1.2,0.3855743885040283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,power_law_1.2,5.749881744384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,power_law_1.2,0.4897280216217041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.09318400025367737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.12246400117874146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.13706239461898803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.1780608057975769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.24227840900421144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.2885119915008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.3618495941162109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.5276800155639648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.7963903903961181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.9494272232055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.07201279997825623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.07924479842185975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.08916479945182801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.10554879903793335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.10819200277328492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.10668799877166749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.13766399621963502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.16997120380401612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.19352320432662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.2555455923080444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.2874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,power_law_1.01,0.40051841735839844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,power_law_1.01,0.5107711791992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.01,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.01,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.01,0.10168319940567017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.01,0.12648320198059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.01,0.1750656008720398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.01,0.18417279720306395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.01,0.1961408019065857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.01,0.19671679735183717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.01,0.18492799997329712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.01,0.19359359741210938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.01,0.20249600410461427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.01,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.01,0.20503039360046388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.01,0.20869760513305663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.01,0.21365759372711182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.01,0.23793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.01,0.264684796333313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.01,0.25477759838104247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.01,0.2957184076309204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.01,0.3538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.01,0.4003136157989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.01,0.4269567966461182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.01,0.506220817565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.01,0.6542272090911865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.01,0.8120639801025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.01,1.0158464431762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,power_law_1.01,0.7108672142028809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,power_law_1.01,0.909939193725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.2,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.2,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.2,0.09834240078926086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.2,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.2,0.1508031964302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.2,0.21777279376983644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.2,0.21905920505523682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.2,0.23323519229888917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.2,0.2348736047744751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.2,0.24774401187896727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.2,0.25290238857269287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.2,0.2568511962890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.2,0.2614464044570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.2,0.27224318981170653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.2,0.2888511896133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.2,0.2876032114028931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.2,0.32378880977630614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.2,0.3709696054458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.2,0.3929152011871338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.2,0.48752641677856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.2,0.5892479896545411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.2,0.7348288059234619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.2,0.8394880294799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.2,1.1301055908203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.2,1.4411199569702149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.2,1.982579231262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.2,2.5451520919799804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,balanced,0.06251200040181477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,balanced,0.11538133025169373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,balanced,0.18127999703089395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,balanced,0.3125813404719035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,balanced,0.31355200211207074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,balanced,0.31571199496587116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,balanced,0.31876800457636517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,balanced,0.32210665941238403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,balanced,0.3240480025609334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,balanced,0.33159466584523517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,balanced,0.3320693373680115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,balanced,0.3363786538441976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,balanced,0.34732266267140705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,balanced,0.35195199648539227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,balanced,0.362554669380188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,balanced,0.3904106616973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,balanced,0.41227201620737713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,balanced,0.466106653213501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,balanced,0.5292426745096842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,balanced,0.6624533335367838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,balanced,0.7515253225962321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,balanced,1.0417546431223552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,balanced,1.2367733319600422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,balanced,1.7857866287231445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,balanced,2.238277276357015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,balanced,0.07670933504899342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,balanced,0.08055466910203297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,balanced,0.10577066739400227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,balanced,0.14589866995811462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,balanced,0.2116746703783671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,balanced,0.21021332343419394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,balanced,0.21421333154042563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,balanced,0.2111626664797465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,balanced,0.2140000065167745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,balanced,0.2180053393046061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,balanced,0.22219200929005942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,balanced,0.2243679960568746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,balanced,0.23149865865707397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,balanced,0.2402613361676534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,balanced,0.24571200211842856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,balanced,0.25943466027577716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,balanced,0.28867199023564655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,balanced,0.31709333260854083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,balanced,0.3796693483988444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,balanced,0.4341919819513957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,balanced,0.5530879894892374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,balanced,0.6705066363016764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,balanced,0.9705599943796793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,balanced,1.2084533373514812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,balanced,1.7646560668945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,balanced,2.3105386098225913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,power_law_1.2,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,power_law_1.2,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,power_law_1.2,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,power_law_1.2,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,power_law_1.2,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,power_law_1.2,0.085971200466156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,power_law_1.2,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,power_law_1.2,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,power_law_1.2,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,power_law_1.2,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,power_law_1.2,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,power_law_1.2,0.12294399738311768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,power_law_1.2,0.15537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,power_law_1.2,0.17882239818572998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,power_law_1.2,0.24192640781402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,power_law_1.2,0.31251840591430663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,power_law_1.2,0.46382718086242675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,power_law_1.2,0.5981823921203613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,power_law_1.2,0.896883201599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,power_law_1.2,1.1557248115539551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.15129599571228028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.24721920490264893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.43430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.813260841369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.23189115524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.5995391845703124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.2761472702026366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.4069503784179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.4865087509155273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.497248077392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.569158363342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,power_law_1.01,2.6684288024902343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,power_law_1.01,2.7171648025512694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,power_law_1.01,2.7784704208374023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,power_law_1.01,2.881023979187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,power_law_1.01,2.9198272705078123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.0077951431274412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.259008026123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,power_law_1.01,3.355174255371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,power_law_1.01,3.676639938354492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,power_law_1.01,3.659622573852539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,power_law_1.01,4.106284713745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,power_law_1.01,4.47918701171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,power_law_1.01,5.267558288574219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,power_law_1.01,5.81446418762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,power_law_1.01,7.410483551025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,power_law_1.01,8.983052825927734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,balanced,0.0788320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,balanced,0.10723732908566792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,balanced,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,balanced,0.1571999986966451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,balanced,0.1575146714846293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,balanced,0.15611732999483743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,balanced,0.1578986644744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,balanced,0.15730133652687073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,balanced,0.15898133317629495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,balanced,0.16204266746838888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,balanced,0.16235733032226562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,balanced,0.16462399562199911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,balanced,0.16902933518091837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,balanced,0.17138665914535522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,balanced,0.17647467056910196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,balanced,0.1999946633974711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,balanced,0.19969600439071655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,balanced,0.24799466133117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,balanced,0.2487786610921224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,balanced,0.36773331960042316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,balanced,0.32288533449172974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,balanced,0.4551253318786621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,balanced,0.4733866850535075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,balanced,0.686352014541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,balanced,0.7693866888682047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.07790719866752624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08657280206680298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.09341440200805665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.10328960418701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.1292736053466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.15470720529556276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.17051520347595214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.2189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.27790079116821287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.3603071928024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.447219181060791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.6335807800292969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.80764799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07804160118103028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.09242240190505982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.08995199799537659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.1159168004989624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.13525760173797607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.1549183964729309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.16153600215911865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.16568319797515868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.17188479900360107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.17448960542678832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.17804800271987914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.18149759769439697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.17929600477218627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.18910720348358154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.19591679573059081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.20893440246582032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.2346496105194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.24130558967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.29040000438690183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.30869760513305666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.36839680671691893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,power_law_1.01,0.4276224136352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,power_law_1.01,0.5957183837890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,power_law_1.01,0.7007423877716065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,power_law_1.01,0.9996095657348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,power_law_1.01,1.2228032112121583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,power_law_1.2,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,power_law_1.2,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,power_law_1.2,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,power_law_1.2,0.13432960510253905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,power_law_1.2,0.15396480560302733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,power_law_1.2,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,power_law_1.2,0.20081279277801514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,power_law_1.2,0.1995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,power_law_1.2,0.18919680118560792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,power_law_1.2,0.19338239431381227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,power_law_1.2,0.21112959384918212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,power_law_1.2,0.2138495922088623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,power_law_1.2,0.21670401096343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,power_law_1.2,0.21974399089813232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,power_law_1.2,0.22272000312805176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,4,power_law_1.2,0.22369918823242188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,4,power_law_1.2,0.22952959537506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,4,power_law_1.2,0.2422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.2,0.25742080211639407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.2,0.2971839904785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.2,0.33251841068267823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.2,0.3875135898590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.2,0.5020991802215576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.2,0.6593728065490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.2,0.8488896369934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.2,1.2113856315612792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.2,1.5650431632995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.09722880125045777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.09585279822349549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.09996799826622009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.09863680005073547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.09754239916801452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.09779199957847595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.10551040172576905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.10667519569396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.10949120521545411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.11724159717559815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.21498239040374756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.24198400974273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.329203200340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.4330239772796631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.596947193145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.7688384056091309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,balanced,0.05845866600672404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,balanced,0.0533493310213089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,balanced,0.052895997961362205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,balanced,0.05835199852784475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,balanced,0.06041066845258077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,balanced,0.06672533353169759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,balanced,0.07478933533032735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,balanced,0.07672533392906189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.10202133655548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.10096533099810283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.13101333379745483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.15356266498565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.20045334100723267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.24580266078313193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.3402506510416667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.4315146605173747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.06115840077400207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.06736000180244446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.12067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.10085120201110839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.10895359516143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.19190399646759032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.23303680419921874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.32953600883483886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.3903808116912842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,power_law_1.01,0.562009620666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,power_law_1.01,0.7785855770111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.14710400104522706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.2042623996734619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.2569920063018799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.3677504062652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.08593279719352723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.13320319652557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.1674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.19886080026626587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.24676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.26017279624938966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.2649471998214722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.2700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.27767679691314695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.2853503942489624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.29112958908081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.30360960960388184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.309497594833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.31863040924072267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.3366847991943359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.38190081119537356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.42238078117370603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.5098368167877197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5854335784912109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.7562111854553223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.9210176467895508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.262502384185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.6298688888549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.3184831619262694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.0624256134033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,power_law_1.2,0.4700352191925049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,power_law_1.01,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,power_law_1.01,0.14472960233688353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,power_law_1.01,0.19483519792556764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,power_law_1.01,0.18851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,power_law_1.01,0.19233920574188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,power_law_1.01,0.1949952006340027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,power_law_1.01,0.1715456008911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,power_law_1.01,0.17227519750595094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,power_law_1.01,0.1708799958229065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,power_law_1.01,0.18198399543762206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,power_law_1.01,0.1892416000366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,power_law_1.01,0.19258879423141478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,power_law_1.01,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,power_law_1.01,0.18008320331573485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,power_law_1.01,0.19998079538345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,power_law_1.01,0.1974527955055237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,power_law_1.01,0.20081279277801514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,power_law_1.01,0.22917120456695556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,power_law_1.01,0.2488192081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,power_law_1.01,0.3011136054992676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,power_law_1.01,0.30877439975738524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,power_law_1.01,0.4034304141998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,power_law_1.01,0.4653312206268311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,power_law_1.01,0.641542387008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,power_law_1.01,0.6848000049591064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,power_law_1.01,0.9346303939819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,power_law_1.01,1.1734784126281739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,balanced,0.036506667733192444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,balanced,0.03920000046491623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.05286933481693268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.057818666100502014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.06445866823196411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.08076799909273784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.09803199768066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.1275200049082438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.14214932918548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.19141866763432822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.24286933739980063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.01,0.13100160360336305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.01,0.1854464054107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.01,0.30460801124572756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.01,0.39854719638824465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.01,0.6474624156951905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.01,0.7011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.01,0.7221888065338135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.01,0.741644811630249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.01,0.7461567878723144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.01,0.7794047832489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.01,0.7881728172302246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.01,0.805900764465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.01,0.8257984161376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.01,0.860371208190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.01,0.8722304344177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.01,0.903104019165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.01,1.0111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.01,1.0695679664611817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.01,1.2531328201293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.01,1.3459327697753907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.01,1.6870655059814452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.01,1.8688127517700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.01,2.4080127716064452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.01,3.080044746398926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.01,4.3667137145996096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.01,5.30316162109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,balanced,0.048207998275756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,balanced,0.05073066552480062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,balanced,0.05983466903368632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,balanced,0.08267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,balanced,0.09971200426419576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,balanced,0.09912000099817912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,balanced,0.12969600160916647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,balanced,0.14174933234850565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,balanced,0.1875093380610148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,balanced,0.18988800048828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,balanced,0.25595200061798096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,balanced,0.2789813280105591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,balanced,0.38019732634226483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,balanced,0.45826133092244464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,power_law_1.2,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,power_law_1.2,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,power_law_1.2,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,power_law_1.2,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,power_law_1.2,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,power_law_1.2,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,power_law_1.2,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,power_law_1.2,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,power_law_1.2,0.13167359828948974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,power_law_1.2,0.15202560424804687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,power_law_1.2,0.21185920238494874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,power_law_1.2,0.25021440982818605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,power_law_1.2,0.35739519596099856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,power_law_1.2,0.5274879932403564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,balanced,0.07036800185839336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,balanced,0.07073600093523662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,balanced,0.08564800024032593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,balanced,0.09680533409118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,balanced,0.1009333332379659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,balanced,0.12203733126322429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,balanced,0.13822399576505026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,balanced,0.17568532625834146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,balanced,0.19777067502339682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,balanced,0.29949333270390827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,balanced,0.3630559841791789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,balanced,0.5401120185852051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,balanced,0.698362668355306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,balanced,1.0441866715749104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,balanced,1.3688853581746419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,balanced,0.029685333371162415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,balanced,0.03390933324893316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,balanced,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,balanced,0.0441599984963735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,balanced,0.05366933345794678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,balanced,0.09387200077374776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,balanced,0.11546132961908977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,balanced,0.14761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,balanced,0.1893440087636312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,balanced,0.2539573311805725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,balanced,0.3165066639582316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,balanced,0.041109333435694374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,balanced,0.059088001648585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,balanced,0.07410133381684621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,balanced,0.07491733133792877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,balanced,0.07569600145022075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,balanced,0.07713599999745686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,balanced,0.07678399980068207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,balanced,0.0780213326215744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,balanced,0.0782239983479182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,balanced,0.07858666777610779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,balanced,0.08409600456555684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,balanced,0.08981866637865703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,balanced,0.09485333164532979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,balanced,0.09887466828028361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,balanced,0.1074720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,balanced,0.11731732885042827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,balanced,0.13545067111651102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,balanced,0.15241600076357523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,balanced,0.18582399686177573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,balanced,0.22587732474009195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,balanced,0.3004639943440755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,balanced,0.36897599697113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,balanced,0.04414933423201243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,balanced,0.06297599772612254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,balanced,0.10496000448862712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,balanced,0.16262933611869812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,balanced,0.21094399690628052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,balanced,0.21336533625920615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,balanced,0.21518399318059286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,balanced,0.21278933684031168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,balanced,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,balanced,0.21548799673716226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,balanced,0.2161066730817159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,balanced,0.21861867109934488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,balanced,0.220960001150767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,balanced,0.22667733828226724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,balanced,0.2286506692568461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,balanced,0.2308853268623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,balanced,0.2644373377164205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,balanced,0.24195200204849243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,balanced,0.2985173265139262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,balanced,0.2749119997024536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,balanced,0.4217439889907837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,balanced,0.3449759880701701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,balanced,0.5296746492385864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,balanced,0.5213226477305094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,balanced,0.7786453564961752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,balanced,0.7689759731292725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,power_law_1.2,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,power_law_1.2,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,power_law_1.2,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,power_law_1.2,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,power_law_1.2,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.1384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,balanced,0.04538666705290476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,balanced,0.04915200173854828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,balanced,0.06418666740258534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,balanced,0.06586666901906331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,balanced,0.07625600198904674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,balanced,0.09482666850090027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,balanced,0.11746133367220561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,balanced,0.13618666927019754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,balanced,0.18147732814153036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,balanced,0.2200373411178589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,balanced,0.30973867575327557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,balanced,0.3875573476155599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.19385600090026855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.14431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.20279040336608886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.24973440170288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.2503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.2405695915222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.27313919067382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.2771903991699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.28003840446472167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.275052809715271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.2929215908050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.29932799339294436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.3145983934402466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.34618239402770995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.3763391971588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.44411520957946776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.49152002334594724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.6060160160064697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.7253824234008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,0.9737088203430175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.1984959602355958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,1.6503231048583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,2.199507141113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,power_law_1.2,0.25841920375823973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,balanced,0.052255998055140175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,balanced,0.07737066845099132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,balanced,0.12876799702644348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,balanced,0.12139200170834859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,balanced,0.12424000104268391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,balanced,0.12572800119717917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,balanced,0.12300800283749898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,balanced,0.12613333264986673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,balanced,0.1276746690273285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,balanced,0.12871467073758444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,balanced,0.13312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,balanced,0.1325920025507609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,balanced,0.1388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,balanced,0.1421066621939341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,balanced,0.14867200454076132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,balanced,0.16062399744987488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,balanced,0.17198934157689413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,balanced,0.20587199926376343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,balanced,0.2230506738026937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,balanced,0.2847040096918742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,balanced,0.3243573307991028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,balanced,0.448202649752299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,balanced,0.5515253146489462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,balanced,0.768069346745809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,balanced,1.0000639756520588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,power_law_1.2,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,power_law_1.2,0.0689408004283905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,power_law_1.2,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,power_law_1.2,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,power_law_1.2,0.0908415973186493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,power_law_1.2,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,power_law_1.2,0.10538239479064941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,power_law_1.2,0.11134079694747925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,power_law_1.2,0.1262336015701294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,power_law_1.2,0.18329600095748902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,power_law_1.2,0.21373438835144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,power_law_1.2,0.2685631990432739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,power_law_1.2,0.34247679710388185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,power_law_1.2,0.47983360290527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,power_law_1.2,0.6209472179412842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,power_law_1.2,0.8709888458251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,power_law_1.2,1.2167488098144532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.02871679961681366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.11250560283660889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.12684160470962524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.16851840019226075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.2053056001663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,power_law_1.01,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,power_law_1.01,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,power_law_1.01,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,power_law_1.01,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,power_law_1.01,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,power_law_1.01,0.08681600093841553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,power_law_1.01,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,power_law_1.01,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,power_law_1.01,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,power_law_1.01,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,power_law_1.01,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,power_law_1.01,0.11143039464950562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,power_law_1.01,0.12067840099334717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,power_law_1.01,0.12723840475082399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,power_law_1.01,0.1720639944076538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,power_law_1.01,0.2045056104660034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,power_law_1.01,0.24483840465545653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.29319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,power_law_1.01,0.29976320266723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,power_law_1.01,0.493120002746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,power_law_1.01,0.5127423763275146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,power_law_1.01,0.7043263912200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.4031040191650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,power_law_1.01,0.863923168182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,power_law_1.01,1.5306431770324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,power_law_1.2,0.06678400039672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,power_law_1.2,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,power_law_1.2,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,power_law_1.2,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,power_law_1.2,0.056953597068786624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,power_law_1.2,0.0630016028881073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,power_law_1.2,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,power_law_1.2,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,power_law_1.2,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,power_law_1.2,0.14060800075531005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,power_law_1.2,0.1740671992301941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,power_law_1.2,0.22130560874938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,power_law_1.2,0.2685823917388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,power_law_1.2,0.39685759544372556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,power_law_1.2,0.4775231838226318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,power_law_1.2,0.706112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,power_law_1.2,0.8784895896911621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,power_law_1.01,1.800396728515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.2,0.04834559857845307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.2,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.2,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.2,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.2,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.2,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.2,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.2,0.13627519607543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.2,0.14839040040969848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.2,0.20803840160369874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.2,0.19757440090179443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.2,0.2743488073348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.2,0.3694783926010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.2,0.5332992076873779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.2,0.5575488090515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,balanced,0.0769706666469574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,balanced,0.11128000418345134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,balanced,0.17397334178288779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,balanced,0.17613333463668823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,balanced,0.17775466044743857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,balanced,0.1771519978841146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,balanced,0.17825599511464438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,balanced,0.17976532379786173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,balanced,0.18178667624791464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,balanced,0.18227199713389078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,balanced,0.18522133429845175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,balanced,0.1938986579577128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,balanced,0.19417067368825278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,balanced,0.19909334182739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,balanced,0.21445866425832114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,balanced,0.22068800528844199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,balanced,0.24388800064722696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,balanced,0.2696320017178853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,balanced,0.3152586619059245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,balanced,0.3473546504974365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,balanced,0.4590666691462199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,balanced,0.5378613471984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,balanced,0.7582453091939291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,balanced,0.9270186424255371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,balanced,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,balanced,0.04663999875386556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,balanced,0.05017066498597463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,balanced,0.0714026689529419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,balanced,0.0728959987560908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,balanced,0.07402133444945018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,balanced,0.07583466668923695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,balanced,0.07629866898059845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,balanced,0.07644799848397572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,balanced,0.07719466586907704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,balanced,0.08384533723195393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,balanced,0.08385066191355388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,balanced,0.0902346670627594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,balanced,0.10283199946085612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,balanced,0.1165066659450531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,balanced,0.12892267107963562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,balanced,0.1553386648495992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,balanced,0.1934986710548401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,balanced,0.2521173357963562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,balanced,0.3022453387578328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,balanced,0.4252266486485799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,balanced,0.5285919904708862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.13004159927368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.16559360027313233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.19673600196838378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.3230783939361572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.34644479751586915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.35844480991363525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.3573375940322876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.37430400848388673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.38541440963745116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.3994175910949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.4083263874053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.42495999336242674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.4356991767883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.454860782623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.47769598960876464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5379392147064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6039872169494629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.7305984020233154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.7881343841552735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,0.9848128318786621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.1784000396728516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,1.5406335830688476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.09688959717750549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.12000000476837158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,1.896076774597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.20334720611572266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.20274560451507567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,2.6092927932739256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.20540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.21843841075897216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.22081279754638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.22602880001068115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.2287168025970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.2411776065826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.25541119575500487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.25235838890075685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.2763711929321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.30972800254821775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.3434943914413452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.4347008228302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.4666175842285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.5760064125061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.7079936027526855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,0.8945088386535645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.1873536109924316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,1.68603515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,3.395577621459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,1.9588863372802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.08925439715385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.10147839784622192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.1315135955810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.14654719829559326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.17031680345535277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.19128960371017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.23749120235443116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.2859328031539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.3755264043807983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,power_law_1.01,0.4670271873474121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,power_law_1.01,0.6507775783538818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,power_law_1.01,0.8385024070739746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,balanced,0.03930133332808813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,balanced,0.0393653338154157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,balanced,0.045514668027559914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.06814399858315785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.08648000160853068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.1053493320941925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.13398399949073792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.1609760026137034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.22506133715311685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.28175467252731323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.04051199853420258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.051974397897720334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.05736320018768311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.06117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.12664320468902587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.1304128050804138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.2
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.2134848117828369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.2720767974853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,power_law_1.2,0.3690432071685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,power_law_1.2,0.49763197898864747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,power_law_1.2,0.6218495845794678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,power_law_1.2,0.026617598533630372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,power_law_1.2,0.02457599937915802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,power_law_1.2,0.028364801406860353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,power_law_1.2,0.026265600323677064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,power_law_1.2,0.026694399118423463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,power_law_1.2,0.02715519964694977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,power_law_1.2,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,power_law_1.2,0.02928000092506409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,power_law_1.2,0.03310079872608185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,power_law_1.2,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,power_law_1.2,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.07987200021743775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,power_law_1.01,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,power_law_1.01,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,power_law_1.01,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,power_law_1.01,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,power_law_1.01,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,power_law_1.01,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,power_law_1.01,0.10747519731521607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,power_law_1.01,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,power_law_1.01,0.1537727952003479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.12958719730377197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,power_law_1.01,0.18287999629974366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.158297598361969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,power_law_1.01,0.24515841007232667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,power_law_1.01,0.32204160690307615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,power_law_1.01,0.45711359977722166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,power_law_1.01,0.5816768169403076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,power_law_1.01,0.8714112281799317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,power_law_1.01,1.096083164215088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.22586240768432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,power_law_1.2,0.2859839916229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,balanced,0.07671999931335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,balanced,0.15496533115704855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,balanced,0.1840320030848185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,balanced,0.18572266896565756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,balanced,0.1870666742324829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,balanced,0.18638400236765543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,balanced,0.18413333098093668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,balanced,0.1850186586380005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,balanced,0.1853813330332438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,balanced,0.18553600708643594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,balanced,0.18845866123835245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,balanced,0.19176000356674194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,balanced,0.19458132982254028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,balanced,0.19675199190775552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,balanced,0.2192266583442688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,balanced,0.206879993279775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,balanced,0.2430986762046814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,balanced,0.2316853404045105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,balanced,0.32049065828323364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,balanced,0.2906986673672994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,balanced,0.44230401515960693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,balanced,0.43032534917195636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,balanced,0.6799200375874838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,balanced,0.7052106857299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.11278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.13740160465240478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.16099200248718262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.21431679725646974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.2905855894088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.44843521118164065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.5693503856658936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.8715519905090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,1.1184767723083495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,power_law_1.2,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,power_law_1.2,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,power_law_1.2,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,power_law_1.2,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,power_law_1.2,0.06849279999732971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,power_law_1.2,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,power_law_1.2,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,power_law_1.2,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,power_law_1.2,0.0847104012966156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,power_law_1.2,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,power_law_1.2,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,power_law_1.2,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,power_law_1.2,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,power_law_1.2,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,power_law_1.2,0.10926719903945922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,power_law_1.2,0.11898880004882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,power_law_1.2,0.17057280540466307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,power_law_1.2,0.1397696018218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.2,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,power_law_1.2,0.19471360445022584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.2,0.2906816005706787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.2,0.3582655906677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.2,0.5533887863159179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.2,0.6678400039672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,balanced,0.0351946676770846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,balanced,0.03316800047953924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,balanced,0.03428266694148382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,balanced,0.0367253323396047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,balanced,0.03523733218510946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,balanced,0.04398400088151296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,balanced,0.05299200117588043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,balanced,0.050250664353370667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,balanced,0.115365336338679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,balanced,0.1555466651916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,balanced,0.18821332852045694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,balanced,0.26160534222920734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,balanced,0.32227200269699097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.2,1.0404224395751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,power_law_1.2,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,power_law_1.2,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,power_law_1.2,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,power_law_1.2,0.0523904025554657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,power_law_1.2,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,power_law_1.2,0.2543423891067505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,power_law_1.2,0.3858815908432007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,power_law_1.2,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.2,0.11013120412826538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.2,0.1534656047821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.2,1.3056256294250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.2,0.1897472023963928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.2,0.2804352045059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.2,0.37754240036010744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.2,2.0227840423583983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.2,0.5556096076965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.2,0.7632256031036377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.2,1.119142436981201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.2,2.664588737487793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.2,1.3989760398864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.08769919872283935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.14613759517669678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.18899199962615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,balanced,0.06373333434263866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,balanced,0.08698667089144389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,balanced,0.11136533816655476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,balanced,0.10937600334485371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,balanced,0.11110933621724446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,balanced,0.11081066727638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,balanced,0.11475732922554016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,balanced,0.11878400047620137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,balanced,0.11934399604797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,balanced,0.11949867010116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,balanced,0.12994666894276938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,balanced,0.12950399518013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,balanced,0.13635733723640442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,balanced,0.1421173314253489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,balanced,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,balanced,0.17034665743509927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,balanced,0.18999467293421426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,balanced,0.2276159922281901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,balanced,0.26868800322214764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,balanced,0.346837321917216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,balanced,0.44438934326171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,balanced,0.5936799844106039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,balanced,0.7652160326639811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,balanced,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,balanced,0.0598826656738917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,balanced,0.0701279987891515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,balanced,0.07864533364772797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,balanced,0.08662399649620056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,balanced,0.10169600447018941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,balanced,0.11954666177431743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,balanced,0.17756799856821695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,balanced,0.22000000874201456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,balanced,0.3038453261057536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,balanced,0.38518933455149335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,power_law_1.2,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,power_law_1.2,0.06399999856948853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,power_law_1.2,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,power_law_1.2,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,power_law_1.2,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,power_law_1.2,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,power_law_1.2,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,power_law_1.2,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,power_law_1.2,0.07781760096549988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,power_law_1.2,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,power_law_1.2,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,power_law_1.2,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,power_law_1.2,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,power_law_1.2,0.15337599515914918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,power_law_1.2,0.17482880353927613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,power_law_1.2,0.22902400493621827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,power_law_1.2,0.27603199481964114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,power_law_1.2,0.4134208202362061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,power_law_1.2,0.5107135772705078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,power_law_1.2,0.7507008075714111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,power_law_1.2,1.067807960510254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,power_law_1.01,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,power_law_1.01,0.024435199797153473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,power_law_1.01,0.023929600417613984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,power_law_1.01,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,power_law_1.01,0.0247871994972229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,power_law_1.01,0.02650879919528961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,power_law_1.01,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,power_law_1.01,0.02746880054473877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,power_law_1.01,0.02911359965801239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,power_law_1.01,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,power_law_1.01,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,power_law_1.01,0.04047360122203827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,power_law_1.01,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.12590080499649048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.17009279727935792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.21082239151000975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,balanced,0.03793599953254064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,balanced,0.040037333965301514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,balanced,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,balanced,0.04651199777921041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,balanced,0.04808000226815542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,balanced,0.0705386648575465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,balanced,0.08080000181992848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,balanced,0.08708799878756206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,balanced,0.11117333173751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,balanced,0.12587199608484903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,balanced,0.16662933429082236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,balanced,0.20544532934824625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,balanced,0.28775999943415326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,balanced,0.3657919963200887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,balanced,0.06445866823196411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,balanced,0.1691360076268514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,balanced,0.2148266633351644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,balanced,0.21582400798797607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,balanced,0.22038400173187256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,balanced,0.2174933354059855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,balanced,0.21854400634765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,balanced,0.22005865971247354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,balanced,0.22102399667104086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,balanced,0.22364266713460287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,balanced,0.22586133082707724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,balanced,0.2299306591351827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,balanced,0.2346986730893453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,balanced,0.23990400632222494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,balanced,0.2895626624425252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,balanced,0.26285332441329956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,balanced,0.3197919925053914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,balanced,0.30474666754404706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,balanced,0.42690134048461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,balanced,0.41673068205515545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,balanced,0.5910506645838419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,balanced,0.6342346668243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,balanced,0.9132479826609293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,balanced,1.074186642964681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,balanced,0.050928001602490745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,balanced,0.07054933408896129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,balanced,0.09617066383361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,balanced,0.14802133043607077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,balanced,0.24256000916163126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,balanced,0.24444266160329184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,balanced,0.24490133921305338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,balanced,0.24869867165883383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,balanced,0.24632000923156738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,balanced,0.2512106696764628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,balanced,0.25167999664942425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,balanced,0.25641600290934247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,balanced,0.25340267022450763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,balanced,0.2643360098203023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,balanced,0.2646186749140422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,balanced,0.27085334062576294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,balanced,0.28515734275182086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,balanced,0.2967093388239543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,balanced,0.32517866293589276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,balanced,0.3561546802520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,balanced,0.3988320032755534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,balanced,0.4486773411432902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,balanced,0.5753920078277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,balanced,0.6546080112457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,balanced,0.8904159863789877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,balanced,1.0966026782989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,power_law_1.01,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,power_law_1.01,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,power_law_1.01,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,power_law_1.01,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,power_law_1.01,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,power_law_1.01,0.08296319842338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,power_law_1.01,0.10405759811401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,power_law_1.01,0.10407040119171143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,power_law_1.01,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,power_law_1.01,0.16254080533981324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.01,0.18803199529647827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.01,0.25474560260772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.01,0.28317439556121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.01,0.44494080543518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.01,0.5406720161437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.01,0.758348798751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.01,1.0530943870544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.01,1.671513557434082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.01,2.249401664733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,balanced,0.04452266792456309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,balanced,0.08268799881140391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,balanced,0.08539733290672302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,balanced,0.08669333656628926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,balanced,0.08958933750788371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,balanced,0.08828266461690266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,balanced,0.09168533484141032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,balanced,0.09316266576449077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,balanced,0.10090133547782898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,balanced,0.103301336367925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,balanced,0.10558933019638062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,balanced,0.11756267150243123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,balanced,0.12678399682044983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,balanced,0.1495413382848104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,balanced,0.16756800810496011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,balanced,0.22031466166178384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,balanced,0.2518453399340312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,balanced,0.37060801188151044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,balanced,0.4171573321024577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,balanced,0.5755893389383951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,balanced,0.7521440188090006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.01,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.01,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.01,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.01,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.01,0.0657472014427185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.01,0.08461440205574036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.01,0.09225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.01,0.11367039680480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.01,0.14171520471572877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.01,0.18303999900817872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.01,0.18316160440444945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.01,0.23788158893585204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.01,0.2861056089401245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.01,0.3893631935119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.01,0.4698495864868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,power_law_1.2,0.09658880233764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,power_law_1.2,0.11612160205841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,power_law_1.2,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,power_law_1.2,0.13979519605636598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,power_law_1.2,0.16348799467086791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,power_law_1.2,0.16641279458999633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,power_law_1.2,0.1686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,power_law_1.2,0.17744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,power_law_1.2,0.1716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,power_law_1.2,0.1868288040161133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,power_law_1.2,0.1933632016181946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,power_law_1.2,0.20387840270996094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,power_law_1.2,0.20586879253387452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,power_law_1.2,0.22108159065246583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,power_law_1.2,0.24243199825286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.2,0.27192959785461424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.2,0.32702078819274905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.2,0.37738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.2,0.4843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.2,0.6036736011505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.2,0.8296128273010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.2,1.034995174407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.2,1.4557120323181152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.2,1.955673599243164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,power_law_1.2,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,power_law_1.2,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,power_law_1.2,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,power_law_1.2,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,power_law_1.2,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,power_law_1.2,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,power_law_1.2,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,power_law_1.2,0.07810559868812561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.2,0.1316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,balanced,0.056688000758488975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,balanced,0.07932266592979431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,balanced,0.11308800180753072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,balanced,0.17062934239705405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,balanced,0.20040533939997354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,balanced,0.20633065700531006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,balanced,0.20773865779240927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,balanced,0.20945600668589273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,balanced,0.21145067612330118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,balanced,0.2156533400217692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,balanced,0.21917333205540976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,balanced,0.22583999236424765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,balanced,0.23348800341288248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,balanced,0.23945067326227823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,balanced,0.25249600410461426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,balanced,0.26242132981618244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,balanced,0.2986560066541036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,balanced,0.3227893312772115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,balanced,0.4275519847869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,balanced,0.45153601964314777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,balanced,0.6608373324076334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,balanced,0.7322346369425455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,balanced,1.1474133332570393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,balanced,1.3045813242594402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.2,0.1701696038246155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,balanced,2.1134613355000815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,balanced,2.4406933784484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.2,0.2090751886367798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.2,0.2827840089797974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.2,0.41655678749084474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.2,0.5790080070495606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.2,0.8571455955505372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.2,0.9923520088195801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,balanced,0.05746666590372721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,balanced,0.07979199786980946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,balanced,0.11994666854540507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,balanced,0.12170666456222534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,balanced,0.12081066767374675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,balanced,0.12107200423876445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,balanced,0.12204800049463908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,balanced,0.12120532989501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,balanced,0.1206719974676768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,balanced,0.12229866782824199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,balanced,0.12405866384506226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,balanced,0.1255626678466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,balanced,0.12980799873669943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,balanced,0.13989333311716715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,balanced,0.14312000075976053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,balanced,0.14360533157984415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,balanced,0.17780800660451254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,balanced,0.18054932355880737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,balanced,0.2238933245340983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,balanced,0.2508693337440491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,balanced,0.32755200068155926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,balanced,0.38389333089192706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,balanced,0.543066660563151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,balanced,0.6905972957611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,balanced,1.003216028213501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,balanced,1.300112009048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,balanced,0.06467199822266896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,balanced,0.06782933572928111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,balanced,0.06916800141334534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,balanced,0.06881600121657054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,balanced,0.06829333305358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,balanced,0.07436266541481018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,balanced,0.07435733576615651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,balanced,0.07610133290290833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,balanced,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,balanced,0.08984000484148662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,balanced,0.10218133529027303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,balanced,0.1072160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,balanced,0.11528533697128296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,balanced,0.1218239963054657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,balanced,0.14020267128944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,balanced,0.16211733222007751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,balanced,0.19430933396021524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,balanced,0.236735999584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,balanced,0.3486773173014323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,balanced,0.42723198731740314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,balanced,0.631493330001831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,balanced,0.8147839705149332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,balanced,0.05624000231424967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,balanced,0.06301866471767426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,balanced,0.07239466905593872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.0754559983809789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.10270933310190837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.12431466579437256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.1462506651878357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.18593066930770874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.21453867355982462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.2941173315048218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.36141331990559894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,power_law_1.01,0.07891200184822082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,power_law_1.01,0.10697599649429321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,power_law_1.01,0.12409600019454955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,power_law_1.01,0.16229759454727172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,power_law_1.01,0.19549440145492553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,power_law_1.01,0.22151679992675782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,power_law_1.01,0.2637311935424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,power_law_1.01,0.29818239212036135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,power_law_1.01,0.27521278858184817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,power_law_1.01,0.30868480205535886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,power_law_1.01,0.31872000694274905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,power_law_1.01,0.3199552059173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,power_law_1.01,0.3315648078918457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,power_law_1.01,0.3256063938140869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,power_law_1.01,0.3439039945602417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,power_law_1.01,0.35834879875183107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,power_law_1.01,0.3997056007385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,power_law_1.01,0.432966423034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,power_law_1.01,0.4923583984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,power_law_1.01,0.5497087955474853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,power_law_1.01,0.6921792030334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,power_law_1.01,0.8819392204284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,power_law_1.01,1.1345600128173827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,power_law_1.01,1.336723232269287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,power_law_1.01,1.7950464248657227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,power_law_1.01,2.3
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.1044543981552124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.14675840139389038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.17445119619369506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.21334400177001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.3014336109161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.412934398651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.5535488128662109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,power_law_1.01,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,power_law_1.01,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,power_law_1.01,0.08136320114135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,power_law_1.01,0.09622399806976319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,power_law_1.01,0.09630720019340515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,power_law_1.01,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,power_law_1.01,0.10085120201110839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,power_law_1.01,0.10588799715042115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,power_law_1.01,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,power_law_1.01,0.1239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,power_law_1.01,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,power_law_1.01,0.16921600103378295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,power_law_1.01,0.23471999168395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.01,0.2703295946121216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.01,0.34676480293273926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.01,0.49717121124267577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.01,0.5992063999176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,0.8254528045654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.01,0.8465408325195313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.01,1.2506048202514648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.01,1.8688447952270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.01,2.3058624267578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,0.9389760017395019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.01,2.8177663803100588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,balanced,0.06382399797439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,balanced,0.08448533217112224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,balanced,0.12211733063062032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,balanced,0.19101866086324057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,balanced,0.2236106594403585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,balanced,0.22669333219528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,balanced,0.22790400187174478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,balanced,0.23146667083104452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,balanced,0.2313279906908671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,balanced,0.23458133141199747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,balanced,0.23725332816441855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,balanced,0.23734400669733682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,balanced,0.24374399582544962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,balanced,0.25222400824228924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,balanced,0.25522667169570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,balanced,0.2642293373743693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,balanced,0.286954661210378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.3074293335278829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4334719975789388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.3893066644668579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.5391466617584229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.5827733278274536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.8351786931355795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.9891466299692789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.4120532671610515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.8065706888834636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,balanced,0.0783733328183492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,balanced,0.08075200021266937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,balanced,0.08227733274300893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,balanced,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,balanced,0.08124266564846039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,balanced,0.08236800134181976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,balanced,0.08589866757392883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,balanced,0.08598933617273967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,balanced,0.08604266246159871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,balanced,0.08498133222262065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,balanced,0.08604799707730611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,balanced,0.10500267148017883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,balanced,0.1016533374786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,balanced,0.11534933249155681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,balanced,0.11979732910792033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,balanced,0.14612799882888794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,balanced,0.16079466541608176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,balanced,0.21237866083780924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,balanced,0.25198932488759357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,balanced,0.34804801146189374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,balanced,0.434005339940389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,balanced,0.6321173508961996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,balanced,0.8189547061920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,balanced,0.07735999921957652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,balanced,0.075354665517807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,balanced,0.0766186664501826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,balanced,0.07771733403205872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,balanced,0.07691200077533722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,balanced,0.07690133154392242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,balanced,0.07868800063927968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,balanced,0.08051733175913493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,balanced,0.0804746647675832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,balanced,0.08380800485610962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,balanced,0.08942932883898418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,balanced,0.09106666843096416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,balanced,0.0995253324508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,balanced,0.10462933778762817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,balanced,0.11565333604812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,balanced,0.1360266705354055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,balanced,0.16134400169054666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,balanced,0.20668266216913858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,balanced,0.2474720080693563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,balanced,0.33240000406901044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,balanced,0.3989439805348714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,balanced,0.5818826754887899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,balanced,0.7423199812571207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,power_law_1.2,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,power_law_1.2,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,power_law_1.2,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,power_law_1.2,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,power_law_1.2,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,power_law_1.2,0.06966400146484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,power_law_1.2,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,power_law_1.2,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,power_law_1.2,0.12089600563049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,power_law_1.2,0.14960639476776122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,power_law_1.2,0.18293119668960572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,power_law_1.2,0.2153088092803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,power_law_1.2,0.30359039306640623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,power_law_1.2,0.38625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,power_law_1.2,0.5808063983917237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,power_law_1.2,0.692409610748291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.08812159895896912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.08932480216026306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.1058303952217102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.10842880010604858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.10027519464492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.11159039735794067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.11683199405670167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.12791039943695068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.1325824022293091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.14714239835739135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.19850239753723145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.22383999824523926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.30046079158782957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.3813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.5233727931976319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.6247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,power_law_1.01,0.882636833190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,balanced,0.033914667864640556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.1629440307617187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,balanced,0.0372533326347669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,balanced,0.050800000627835594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,balanced,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,balanced,0.0879253347714742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,balanced,0.09823999802271526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,balanced,0.12967999776204428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,balanced,0.14614400267601013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,balanced,0.20796799659729004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,balanced,0.24413333336512247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,balanced,0.07228266696135204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,balanced,0.10032533605893452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,balanced,0.1534773310025533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,balanced,0.25362666447957355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,balanced,0.25407467285792035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,balanced,0.256117324034373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,balanced,0.2563786705334981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,balanced,0.25967999299367267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,balanced,0.26027733087539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,balanced,0.263866662979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,balanced,0.2646346688270569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,balanced,0.27246934175491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,balanced,0.28060799837112427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,balanced,0.2853333353996277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,balanced,0.29601067304611206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,balanced,0.3171360095342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,balanced,0.33717334270477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,balanced,0.37972267468770343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,balanced,0.41492799917856854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,balanced,0.5187306801478068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,balanced,0.5902986526489258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,balanced,0.8086826801300049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,balanced,0.9465280373891195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,balanced,1.3167466322580974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,balanced,1.6510772705078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,power_law_1.2,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,power_law_1.2,0.07333120107650756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,power_law_1.2,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,power_law_1.2,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,power_law_1.2,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,power_law_1.2,0.09509119987487794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,power_law_1.2,0.09512320160865784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,power_law_1.2,0.09690240025520325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,power_law_1.2,0.09644160270690919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,power_law_1.2,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,power_law_1.2,0.10299520492553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,power_law_1.2,0.10545920133590699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,power_law_1.2,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,power_law_1.2,0.11043839454650879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,power_law_1.2,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.2,0.13180160522460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.2,0.1578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.2,0.18918399810791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.2,0.22828800678253175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.2,0.2761152029037476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.2,0.38807680606842043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.2,0.5133823871612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.2,0.8000576019287109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.2,1.1219136238098144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,balanced,0.0393653338154157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,balanced,0.037920000652472176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,balanced,0.051914667089780174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,balanced,0.05299200117588043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,balanced,0.08888000249862671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,balanced,0.09675733248392741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,balanced,0.14830933014551798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,power_law_1.01,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,power_law_1.01,0.04338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,power_law_1.01,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,power_law_1.01,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,power_law_1.01,0.1181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,power_law_1.01,0.12469760179519654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,power_law_1.01,0.1338368058204651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,power_law_1.01,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,power_law_1.01,0.13864320516586304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,power_law_1.01,0.14606720209121704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,power_law_1.01,0.1506495952606201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,power_law_1.01,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,power_law_1.01,0.16745599508285522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,power_law_1.01,0.1764799952507019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,power_law_1.01,0.18851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,power_law_1.01,0.21719040870666503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,power_law_1.01,0.2525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,power_law_1.01,0.31755518913269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,power_law_1.01,0.3617216110229492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,power_law_1.01,0.4844480037689209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,power_law_1.01,0.5965568065643311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,power_law_1.01,0.8301504135131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,power_law_1.01,1.0479424476623536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,power_law_1.01,1.4921536445617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,power_law_1.01,1.9725055694580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,power_law_1.2,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,power_law_1.2,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,power_law_1.2,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,power_law_1.2,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,power_law_1.2,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,power_law_1.2,0.06849279999732971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,power_law_1.2,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,power_law_1.2,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,power_law_1.2,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,power_law_1.2,0.08758400082588196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,power_law_1.2,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,power_law_1.2,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,power_law_1.2,0.1186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,power_law_1.2,0.14618879556655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,power_law_1.2,0.16371840238571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,power_law_1.2,0.21825919151306153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,power_law_1.2,0.2666368007659912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,power_law_1.2,0.38490240573883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,power_law_1.2,0.475270414352417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,power_law_1.2,0.6979519844055175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,power_law_1.2,0.9441087722778321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.15942399501800536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.18881280422210694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.2457024097442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.25179519653320315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.2662719964981079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.2672895908355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.2764672040939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.2873408079147339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.2946176052093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.3020927906036377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.31363840103149415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.3237567901611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.34495999813079836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.3894207954406738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.4285696029663086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.519865608215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.5943808078765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.7649792194366455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,0.9260992050170899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.2557056427001954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,1.6545215606689454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,2.3458879470825194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,3.0853248596191407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,power_law_1.2,0.10682879686355591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,power_law_1.2,0.12622720003128052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,power_law_1.2,0.15843839645385743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,power_law_1.2,0.20485119819641112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,power_law_1.2,0.20988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,power_law_1.2,0.22349441051483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,power_law_1.2,0.2195136070251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,power_law_1.2,0.22904319763183595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,power_law_1.2,0.22999041080474852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,power_law_1.2,0.2437567949295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,power_law_1.2,0.25841920375823973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,power_law_1.2,0.27583999633789064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,power_law_1.2,0.28183679580688475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,power_law_1.2,0.29383680820465086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,power_law_1.2,0.3274496078491211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,power_law_1.2,0.36496639251708984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,power_law_1.2,0.4256703853607178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,power_law_1.2,0.49565439224243163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,power_law_1.2,0.6220863819122314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,power_law_1.2,0.7285312175750732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,power_law_1.2,0.9883456230163574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,power_law_1.2,1.2128064155578613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,power_law_1.2,1.6784832000732421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,power_law_1.2,2.205407905578613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.08693119883537292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.1463487982749939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.18009599447250366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.18935680389404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.21100800037384032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.22019200325012206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.21251840591430665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.21948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.21400959491729737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.2310784101486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.2263040065765381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.23613440990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.2546623945236206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.2517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.28205440044403074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.31115520000457764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.351910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.401964807510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.5037248134613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.5610047817230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,0.7556992053985596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,0.894598388671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.2213631629943849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,1.4989760398864747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,power_law_1.2,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,power_law_1.2,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,power_law_1.2,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,power_law_1.2,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,power_law_1.2,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,power_law_1.2,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,power_law_1.2,0.09190400242805481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,power_law_1.2,0.10230400562286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,power_law_1.2,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,power_law_1.2,0.11217279434204101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,power_law_1.2,0.12983039617538453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,power_law_1.2,0.15738879442214965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,power_law_1.2,0.18537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,power_law_1.2,0.2412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,power_law_1.2,0.2956671953201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,power_law_1.2,0.4004096031188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,power_law_1.2,0.5047103881835937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,power_law_1.2,0.7059455871582031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,power_law_1.2,0.9237312316894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,power_law_1.2,1.3390015602111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,power_law_1.2,1.7468992233276368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.07443199753761291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.08479359745979309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.08505600094795226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.09041919708251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.11296639442443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.05800319910049438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.15249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.17304960489273072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.22085120677947997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.13025280237197875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.25448319911956785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,power_law_1.2,0.3453183889389038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,power_law_1.2,0.4681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.18760960102081298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,power_law_1.2,0.5617792129516601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,balanced,0.04474666714668274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,balanced,0.04663466910521189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,balanced,0.053727999329566956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,balanced,0.13109333316485086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,balanced,0.1323946714401245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,balanced,0.1318986713886261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,balanced,0.13477866848309836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,balanced,0.13675199945767721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,balanced,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,balanced,0.14199999968210855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,balanced,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,balanced,0.14900267124176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,balanced,0.153519997994105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,balanced,0.15949333707491556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,balanced,0.17192000150680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,balanced,0.1852160096168518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,balanced,0.21039466063181558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,balanced,0.24065599838892618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,balanced,0.2887466748555501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,balanced,0.3485493262608846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,balanced,0.45590933163960773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,balanced,0.5609546502431234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,balanced,0.7746453285217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,balanced,0.9903146425882975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.23057920932769777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,power_law_1.2,0.9350720405578613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,power_law_1.2,1.1668031692504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,power_law_1.2,1.3881728172302246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.07029759883880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.09711999893188476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.1165503978729248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.12147200107574463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.14692480564117433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.1856511950492859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.24830079078674316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.27355520725250243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.4175424098968506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.658406400680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,power_law_1.2,0.05123199820518494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,power_law_1.2,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,power_law_1.2,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,power_law_1.2,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,power_law_1.2,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,power_law_1.2,0.06932479739189149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,power_law_1.2,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.2,0.12382080554962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.2,0.1383039951324463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.2,0.17633919715881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.2,0.2131903886795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.2,0.32391040325164794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.2,0.3607935905456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.2,0.5654463768005371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.2,0.6856063842773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.10093439817428589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.10971519947052003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.11734399795532227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.14110080003738404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.1775231957435608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.20817279815673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.25635199546813964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.33631999492645265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.4253824234008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,power_law_1.01,0.6202047824859619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,power_law_1.01,0.7679488182067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,power_law_1.01,0.10924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,power_law_1.01,0.1287168025970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,power_law_1.01,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,power_law_1.01,0.15828479528427125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,power_law_1.01,0.15678080320358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,power_law_1.01,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,power_law_1.01,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,power_law_1.01,0.1609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,power_law_1.01,0.1549183964729309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,power_law_1.01,0.1665343999862671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,power_law_1.01,0.1630784034729004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,power_law_1.01,0.1710528016090393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,power_law_1.01,0.17455359697341918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,power_law_1.01,0.1870527982711792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,power_law_1.01,0.20280959606170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,power_law_1.01,0.2043071985244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,power_law_1.01,0.22999680042266846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,power_law_1.01,0.2457024097442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,power_law_1.01,0.2993664026260376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,power_law_1.01,0.31616640090942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,power_law_1.01,0.4092735767364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,power_law_1.01,0.4658304214477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,power_law_1.01,0.6406911849975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,power_law_1.01,0.8229375839233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.07022079825401306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.08640000224113464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.11200000047683716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.13238400220870972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.15627520084381102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.19233280420303345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.24447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.31712639331817627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.3955840110778809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,power_law_1.2,0.5658751964569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,power_law_1.2,0.7314752101898193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.19979519844055177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.3199615955352783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.5855103969573975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.2,0.9327872276306153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.2,1.31461763381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.2,2.333964729309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.2,2.5458240509033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.2,2.643270492553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.2,2.6848896026611326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.2,2.7435455322265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.2,2.8517759323120115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.2,2.958624076843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.2,2.9755008697509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.2,3.128870391845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.2,3.209868621826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.2,3.236838531494141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.2,3.4324672698974608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.2,3.6499454498291017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.2,3.823980712890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.2,4.261088180541992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.2,4.219232177734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.2,4.918374252319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.2,5.463334274291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.2,6.531046295166016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.2,8.838047790527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.2,11.341171264648438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.2,12.461939239501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.1500991940498352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.15815039873123168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.16940159797668458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.18929280042648317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.14039679765701293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.1655743956565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.1717311978340149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.1688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.17493760585784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.16785279512405396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.16371840238571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.1831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.18551679849624633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.20131840705871581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.19960960149765014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.23087360858917236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.28906879425048826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.2702591896057129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.32690560817718506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.3718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,power_law_1.2,0.48516478538513186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,power_law_1.2,0.47706241607666017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,power_law_1.2,0.6659200191497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,power_law_1.2,0.7915584087371826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,power_law_1.2,1.1946880340576171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,power_law_1.2,1.6428287506103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.07557119727134705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.10791679620742797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.17146240472793578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.19140479564666749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.18938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.19173760414123536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.19981440305709838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.19634560346603394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.20008320808410646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.2082047939300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.21643519401550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.22539520263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.23635199069976806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.2335103988647461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.25893120765686034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.2832767963409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.32858240604400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.35765120983123777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.4431424140930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,power_law_1.2,0.4987648010253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,power_law_1.2,0.6432767868041992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,power_law_1.2,0.8432703971862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,power_law_1.2,1.119814395904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,power_law_1.2,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,power_law_1.2,0.04371199905872345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,power_law_1.2,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,power_law_1.2,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,power_law_1.2,0.06384639739990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,power_law_1.2,0.07397760152816772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,power_law_1.2,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,power_law_1.2,0.12540800571441652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,power_law_1.2,0.15927679538726808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,power_law_1.2,0.21283199787139892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,power_law_1.2,0.321235203742981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,power_law_1.2,0.4241343975067139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,power_law_1.2,0.6829951763153076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,power_law_1.2,1.4065024375915527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,power_law_1.2,0.867244815826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,power_law_1.2,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,power_law_1.2,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,power_law_1.2,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,power_law_1.2,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,power_law_1.2,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,power_law_1.2,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,power_law_1.2,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,power_law_1.2,0.09947519898414611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,power_law_1.2,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,power_law_1.2,0.13423360586166383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,power_law_1.2,0.15063040256500243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,power_law_1.2,0.2097856044769287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,power_law_1.2,0.2510335922241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,power_law_1.2,0.35555200576782225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,power_law_1.2,0.4680319786071777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,power_law_1.2,0.7062208175659179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,power_law_1.2,0.8215423583984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,balanced,0.032501332461833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,balanced,0.029653333127498627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,balanced,0.03181866556406021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,balanced,0.03279466678698858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,balanced,0.031845333675543465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,balanced,0.0345920001467069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,balanced,0.03589333345492681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,balanced,0.0710506687561671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,balanced,0.11314666271209717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,balanced,0.12490133444468181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,power_law_1.01,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,power_law_1.01,0.05069440007209778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,power_law_1.01,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,power_law_1.01,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,power_law_1.01,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,power_law_1.01,0.06506239771842956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,power_law_1.01,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.01,0.14247679710388184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.01,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.01,0.21765758991241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.01,0.25520639419555663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.01,0.3519615888595581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.01,0.43979520797729493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,2,power_law_1.01,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,2,power_law_1.01,0.14428160190582276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,2,power_law_1.01,0.15618560314178467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,2,power_law_1.01,0.21100800037384032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,2,power_law_1.01,0.2514944076538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,2,power_law_1.01,0.34151039123535154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,2,power_law_1.01,0.3380928039550781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,2,power_law_1.01,0.36290559768676756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,2,power_law_1.01,0.36722559928894044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,2,power_law_1.01,0.3776256084442139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,2,power_law_1.01,0.37903358936309817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,2,power_law_1.01,0.3872704029083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,2,power_law_1.01,0.3891968011856079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,2,power_law_1.01,0.39375360012054444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,2,power_law_1.01,0.4132095813751221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,2,power_law_1.01,0.4142464160919189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,2,power_law_1.01,0.427564811706543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,2,power_law_1.01,0.4570623874664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.01,0.49153919219970704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.01,0.5721792221069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.01,0.6599616050720215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.01,0.8529088020324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.01,0.9891712188720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.01,1.4409024238586425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.01,1.7635135650634766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.01,2.4450687408447265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.01,3.3675521850585937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,power_law_1.01,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,power_law_1.01,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,power_law_1.01,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,power_law_1.01,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,power_law_1.01,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,power_law_1.01,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,power_law_1.01,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,power_law_1.01,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,power_law_1.01,0.15111039876937865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,power_law_1.01,0.20446081161499025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,power_law_1.01,0.2418816089630127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,power_law_1.01,0.35790719985961916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,power_law_1.01,0.4455615997314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,balanced,0.03997866561015447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,balanced,0.03850133220354716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,balanced,0.04119466741879781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,balanced,0.04072533299525579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,balanced,0.04514666895071665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,balanced,0.046944002310434975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.07649600009123485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.08644800384839375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.1122773289680481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.12151466806729634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.15842666228612265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.1927786668141683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,balanced,0.07366933425267537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,balanced,0.09200533231099446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,balanced,0.1384213368097941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,balanced,0.20333333810170492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,balanced,0.25334932406743366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,balanced,0.2531413237253825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,balanced,0.25570134321848553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,balanced,0.2546079953511556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,balanced,0.2577333251635234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,balanced,0.2612160046895345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,balanced,0.26128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,balanced,0.26518932978312176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,balanced,0.2650826573371887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,balanced,0.27183467149734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,balanced,0.2752319971720378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,balanced,0.28119999170303345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,balanced,0.334666649500529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,balanced,0.3116213281949361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,balanced,0.3927946488062541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,balanced,0.37146135171254474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,balanced,0.5785866578420004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,balanced,0.4781759977340698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,balanced,0.705893357594808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,balanced,0.7084320386250814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,balanced,1.0366400082906086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,balanced,1.1423786481221516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.14622080326080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.14743679761886597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.14899840354919433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.15089919567108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.15162240266799926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.15508480072021485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.1551743984222412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.16037119626998902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.16310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.17112319469451903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.19404159784317015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.2027519941329956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.2041152000427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.2549439907073975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.2958143949508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.3631616115570068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.4642752170562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.5408576011657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.7883135795593261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.0784000396728515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.4936448097229005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.073324775695801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,2.5497600555419924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.07273600101470948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.08438400030136109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.10543999671936036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.11700479984283448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.14120320081710816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.22272639274597167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.2548543930053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.36557440757751464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.4303232192993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.6115903854370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.8203200340270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,power_law_1.2,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,power_law_1.2,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,power_law_1.2,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,power_law_1.2,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,power_law_1.2,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,power_law_1.2,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,power_law_1.2,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,power_law_1.2,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,power_law_1.2,0.08199679851531982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,power_law_1.2,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,power_law_1.2,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.2,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.2,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,power_law_1.2,0.10420479774475097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.2,0.17342720031738282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.2,0.22458240985870362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.2,0.28780159950256345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.2,0.3541055917739868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.2,0.5028351783752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.2,0.6751423835754394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,power_law_1.2,0.12324479818344117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,power_law_1.2,0.16037119626998902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,power_law_1.2,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.2,0.7747968196868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,power_law_1.2,0.2874880075454712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,balanced,0.05853866537412008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,balanced,0.05818133552869161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,balanced,0.05855466425418854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,balanced,0.05834666887919108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,balanced,0.061941335598627724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,balanced,0.061941335598627724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,balanced,0.0703893353541692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,balanced,0.10306666294733684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,balanced,0.11415466666221619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,balanced,0.1536853313446045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,balanced,0.14841600259145102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,balanced,0.18695465723673502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,balanced,0.21397866805394491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,balanced,0.27562133471171063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,balanced,0.33398401737213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,power_law_1.2,0.36151039600372314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,power_law_1.2,0.4664768218994141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,power_law_1.2,0.6619967937469482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.11841280460357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.1217344045639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.10843520164489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.11031039953231811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.1144063949584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.11918720006942748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.11998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.12755199670791625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.13226879835128785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.13468799591064454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.14788479804992677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.15772160291671752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.18707200288772582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.2047424077987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.24696319103240966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.28003840446472167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.3779776096343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,power_law_1.01,0.45952639579772947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,power_law_1.01,0.6147007942199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,power_law_1.01,0.7736063957214355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.2,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.2,0.11075199842453003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.2,0.13078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,balanced,0.0584799995024999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,balanced,0.06092800199985504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,balanced,0.06049066781997681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,balanced,0.0610346645116806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,balanced,0.06206400195757548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.2,0.14624639749526977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,balanced,0.0764213353395462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,balanced,0.07638399799664815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,balanced,0.0798933357000351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,balanced,0.08736532926559448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,balanced,0.09732266267140706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,balanced,0.11706667145093282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,balanced,0.13023466865221658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,balanced,0.17029867569605509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,balanced,0.21355734268824259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,balanced,0.29039466381073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,balanced,0.36367468039194745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,balanced,0.5220746596654257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,balanced,0.6679466565450033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.2,0.1559872031211853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.2,0.1387071967124939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.2,0.1364351987838745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.2,0.14398080110549927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.2,0.14572800397872926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.2,0.15288959741592406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.2,0.15162880420684816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.2,0.1575103998184204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.2,0.1638208031654358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.2,0.17761919498443604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.2,0.17875200510025024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.2,0.1896512031555176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.2,0.21358718872070312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.2,0.24442880153656005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.2,0.2996095895767212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.2,0.31923201084136965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.2,0.41301760673522947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.2,0.4371200084686279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.2,0.5853055953979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.2,0.7287871837615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.2,1.1192319869995118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.2,1.2872127532958983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.1309440016746521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.15548160076141357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.18266240358352662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.17767679691314697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.17587200403213502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.170905601978302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.17701120376586915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.1794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.19741439819335938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.20118401050567628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.20365440845489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.2197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.2487231969833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.25191679000854494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.31388800144195556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.33886079788208007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.40215039253234863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.5012671947479248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,power_law_1.2,0.6256256103515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,power_law_1.2,0.8015551567077637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,power_law_1.2,1.1356800079345704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,power_law_1.2,1.3796863555908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,balanced,0.05366933345794678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,balanced,0.058090666929880776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,balanced,0.05765866736570994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,balanced,0.05967999994754791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,balanced,0.06061333417892456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,balanced,0.06968533496061961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,balanced,0.07234133283297221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,balanced,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,balanced,0.08252800007661183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,balanced,0.10102400183677673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,balanced,0.14180800318717957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,balanced,0.17381866772969565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,balanced,0.2294293244679769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,balanced,0.26609599590301514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,balanced,0.3743893305460612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,balanced,0.4761066834131877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,power_law_1.01,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,power_law_1.01,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,power_law_1.01,0.07288320064544677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,power_law_1.01,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,power_law_1.01,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,power_law_1.01,0.07736319899559022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,power_law_1.01,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,power_law_1.01,0.07906559705734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,power_law_1.01,0.08853759765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,power_law_1.01,0.09675520062446594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,power_law_1.01,0.11188479661941528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.01,0.12005120515823364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.01,0.17153279781341552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.01,0.20887680053710939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.01,0.27414400577545167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.01,0.39911038875579835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.01,0.4704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.12396800518035889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.01,0.8443072319030762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.01,1.0109824180603026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.13338240385055541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.18954880237579347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.24702079296112062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,power_law_1.2,0.3611135959625244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,2,power_law_1.01,0.19055999517440797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,2,power_law_1.01,0.22092161178588868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,2,power_law_1.01,0.27847681045532224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,2,power_law_1.01,0.3850048065185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,2,power_law_1.01,0.5235968112945557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,2,power_law_1.01,0.5631616115570068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,2,power_law_1.01,0.8110079765319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,2,power_law_1.01,0.817523193359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,2,power_law_1.01,0.8557696342468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,2,power_law_1.01,0.8756352424621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,2,power_law_1.01,0.8844991683959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,2,power_law_1.01,0.9012607574462891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,2,power_law_1.01,0.927667236328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,2,power_law_1.01,0.9191679954528809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,2,power_law_1.01,0.9426048278808594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,2,power_law_1.01,0.9524736404418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,2,power_law_1.01,0.9578304290771484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,2,power_law_1.01,1.0057727813720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.01,1.060934352874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.01,1.1560895919799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.01,1.2719231605529786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.01,1.446681594848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.01,1.750668716430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,power_law_1.2,0.41560959815979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.01,2.2010944366455076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.01,2.810419273376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.01,3.728537750244141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.01,4.694566345214843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,power_law_1.2,0.6232575893402099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,power_law_1.2,0.9056063652038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,power_law_1.2,0.029478400945663452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,power_law_1.2,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,power_law_1.2,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,power_law_1.2,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,power_law_1.2,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,power_law_1.2,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,power_law_1.2,0.07200000286102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,power_law_1.2,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,power_law_1.2,0.11668479442596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,power_law_1.2,0.1293887972831726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,power_law_1.2,0.17077120542526245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,power_law_1.2,0.21257600784301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,power_law_1.2,0.2941375970840454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,power_law_1.2,0.38172159194946287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,power_law_1.2,0.5485439777374268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,power_law_1.2,0.7073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.01,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.01,0.06596480011940002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.01,0.10442240238189697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.01,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.01,0.11301759481430054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.01,0.11058559417724609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.01,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.01,0.11724159717559815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.01,0.1182144045829773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.01,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.01,0.12488960027694702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.01,0.1314687967300415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.01,0.1368127942085266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.01,0.1466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.01,0.1693824052810669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.01,0.15449600219726561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.01,0.1837504029273987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.01,0.2073983907699585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.01,0.24274559020996095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.01,0.2827647924423218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.01,0.3425663948059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.01,0.39202558994293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.01,0.5409599781036377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.01,0.6787263870239257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.08038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.11673599481582642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.1395135998725891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.15679999589920043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.21251199245452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.23328640460968017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.302892804145813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.3807744026184082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,power_law_1.01,0.568287992477417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,power_law_1.01,0.6699391841888428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,balanced,0.05292266607284546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,balanced,0.05858666698137919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,balanced,0.08117866516113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,balanced,0.11156266927719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,balanced,0.1717066764831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,balanced,0.2265333334604899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,balanced,0.22249066829681396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,balanced,0.2244853377342224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,balanced,0.2247520089149475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,balanced,0.22819733619689941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,balanced,0.2342346707979838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,balanced,0.23509333531061807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,balanced,0.23813867568969727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,balanced,0.2502773404121399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,balanced,0.25437867641448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,balanced,0.270416001478831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,balanced,0.302837332089742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,balanced,0.3372693459192912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,balanced,0.3865120013554891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,balanced,0.43060266971588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,balanced,0.5280106862386068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,balanced,0.6397973299026489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,balanced,0.8300693035125732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,balanced,1.0905493100484211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,balanced,1.4719947179158528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,balanced,1.96234130859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,balanced,0.09644800424575806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,balanced,0.09947199622790019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,balanced,0.0988159974416097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,balanced,0.10153599580128987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,balanced,0.10260799527168274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,balanced,0.1023520032564799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,balanced,0.10318400462468465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,balanced,0.10293866197268169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,balanced,0.10571733117103577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,balanced,0.10613866647084554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,balanced,0.1072746713956197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,balanced,0.11380799611409505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,balanced,0.12359467148780823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,balanced,0.12552000085512796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,balanced,0.13173866271972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,balanced,0.15996266404787698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.17603733142217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.2201813260714213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.26648000876108807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.3587626616160075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.44884800910949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.6225706736246744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.765893300374349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.104090690612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.4514880180358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.11164159774780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.14977920055389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.1696128010749817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.2352832078933716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,power_law_1.2,0.2863231897354126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,power_law_1.2,0.4450687885284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,power_law_1.2,0.6011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.06832000017166137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.07237120270729065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.08251519799232483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.10830080509185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.13257600069046022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.1506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,power_law_1.01,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.18389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,power_law_1.01,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,power_law_1.01,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.2435904026031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,power_law_1.01,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.33473920822143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,power_law_1.01,0.12341760396957398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,power_law_1.01,0.17594879865646362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,power_law_1.01,0.17879680395126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,power_law_1.2,0.5880383968353271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,power_law_1.01,0.18969600200653075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,power_law_1.2,0.6969791889190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,power_law_1.01,0.19408639669418334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,power_law_1.01,0.19562239646911622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,power_law_1.01,0.20634880065917968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,power_law_1.01,0.2097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,power_law_1.01,0.21795198917388917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,power_law_1.01,0.22858240604400634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,power_law_1.01,0.24197120666503907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,power_law_1.01,0.2548223972320557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,power_law_1.01,0.2748800039291382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,power_law_1.01,0.31972479820251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,power_law_1.01,0.3545216083526611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,power_law_1.01,0.44544639587402346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,power_law_1.01,0.5224319934844971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,power_law_1.01,0.6948607921600342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,power_law_1.01,0.9105983734130859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,power_law_1.01,1.217516803741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,power_law_1.01,1.5529343605041503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,power_law_1.01,2.156255912780762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,power_law_1.01,2.7731775283813476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,power_law_1.2,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,power_law_1.2,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,power_law_1.2,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,power_law_1.2,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,power_law_1.2,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,power_law_1.2,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,power_law_1.2,0.08014079928398132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,power_law_1.2,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,power_law_1.2,0.121452796459198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,power_law_1.2,0.1383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,power_law_1.2,0.22897279262542725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,power_law_1.2,0.30275840759277345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,balanced,0.04052799940109253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,balanced,0.07141333321730296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,balanced,0.07287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,balanced,0.07317866881688435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,balanced,0.0759626676638921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,balanced,0.07688533266385396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,balanced,0.07770666480064392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,balanced,0.08070399860541026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,balanced,0.08852266271909077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,balanced,0.09429867068926494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,balanced,0.10408533612887065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,balanced,0.10980799794197083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,balanced,0.12997333208719888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,balanced,0.14822399616241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,balanced,0.1918506622314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,balanced,0.22593067089716592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,balanced,0.3107626636823018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,balanced,0.38701868057250977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,balanced,0.56550399462382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,balanced,0.7298399607340494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,power_law_1.2,0.40843520164489744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,power_law_1.2,0.5544511795043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,power_law_1.2,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,power_law_1.2,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,power_law_1.2,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,power_law_1.2,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,power_law_1.2,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,power_law_1.2,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,power_law_1.2,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,power_law_1.2,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,power_law_1.2,0.1057088017463684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,power_law_1.2,0.14924800395965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,power_law_1.2,0.18199679851531983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,power_law_1.2,0.25260798931121825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,power_law_1.2,0.33264000415802003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,power_law_1.2,0.4807551860809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,power_law_1.2,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,power_law_1.2,0.02635520100593567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,power_law_1.2,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,power_law_1.2,0.0567359983921051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,power_law_1.2,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,power_law_1.2,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,power_law_1.2,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.12318719625473022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.12643200159072876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.16198400259017945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.19889919757843016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.28154239654541013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,power_law_1.2,0.6962111949920654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.39368319511413574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.45724802017211913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,power_law_1.2,0.7051904201507568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,power_law_1.01,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,power_law_1.01,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,power_law_1.01,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,power_law_1.01,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,power_law_1.01,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,power_law_1.01,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.01,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.01,0.09105280041694641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.01,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.01,0.17150720357894897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.01,0.19050240516662598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.01,0.27612159252166746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.01,0.3921792030334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.01,0.6298751831054688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.01,0.9118271827697754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,power_law_1.01,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,power_law_1.01,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,power_law_1.01,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,power_law_1.01,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,power_law_1.01,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,power_law_1.01,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,power_law_1.01,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,power_law_1.01,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,power_law_1.01,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,power_law_1.01,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,power_law_1.01,0.0765504002571106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,power_law_1.01,0.10622080564498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,power_law_1.01,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,power_law_1.01,0.1729024052619934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,power_law_1.01,0.20203518867492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,power_law_1.01,0.25848960876464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,power_law_1.01,0.35406720638275146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,power_law_1.01,0.5095871925354004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,balanced,0.02738133321205775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,balanced,0.026362667481104534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,balanced,0.02735999971628189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,balanced,0.029103999336560566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,balanced,0.0310506671667099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,balanced,0.031871999303499855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,balanced,0.031770666440327965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,balanced,0.06366399923960368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,balanced,0.09246933460235596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,balanced,0.10931199789047241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,balanced,0.1516586641470591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,balanced,0.1892533302307129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,power_law_1.01,0.6804096221923828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.2,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.2,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.2,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.2,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.2,0.055302399396896365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.2,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.2,0.06551679968833923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.2,0.077811199426651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.2,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.2,0.1777151942253113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.2,0.2275264024734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.2,0.36362879276275634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.2,0.3550591945648193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,power_law_1.01,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,power_law_1.01,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,power_law_1.01,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,power_law_1.01,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,power_law_1.01,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,power_law_1.01,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,power_law_1.01,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,power_law_1.01,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,power_law_1.01,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,power_law_1.01,0.09379839897155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,power_law_1.01,0.09676799774169922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,power_law_1.01,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,power_law_1.01,0.10730240345001221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,power_law_1.01,0.1080191969871521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,power_law_1.01,0.1239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,power_law_1.01,0.13521280288696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.01,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.01,0.1703871965408325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.01,0.1967039942741394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.01,0.24459519386291503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.01,0.3096127986907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.01,0.4235328197479248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.01,0.5658432006835937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.01,0.8686271667480469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.01,1.0733375549316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,balanced,0.09082133571306865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,balanced,0.12847466270128885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,balanced,0.20253866910934448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,balanced,0.33686399459838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,balanced,0.44817066192626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,balanced,0.45017067591349286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,balanced,0.4519253174463908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,balanced,0.4529813528060913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,balanced,0.45714131991068524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,balanced,0.4562133153279622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,balanced,0.4605493148167928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,balanced,0.46501866976420086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,balanced,0.4673866828282674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,balanced,0.4764053424199422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,balanced,0.4823840061823527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,balanced,0.4915093183517456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,balanced,0.5614026784896851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,balanced,0.5337866544723511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,balanced,0.6975306669871012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,balanced,0.6191093524297079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,balanced,0.9731520016988119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,balanced,0.7795999844868978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,balanced,1.1812106768290203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,balanced,1.1503360271453857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,balanced,1.7257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,balanced,1.8791786829630535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,power_law_1.2,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,power_law_1.2,0.0800383985042572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,power_law_1.2,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,power_law_1.2,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,power_law_1.2,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,power_law_1.2,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,power_law_1.2,0.08470399975776673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,power_law_1.2,0.08510079979896545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,power_law_1.2,0.09564800262451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,power_law_1.2,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,power_law_1.2,0.10341759920120239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,power_law_1.2,0.11697920560836791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,power_law_1.2,0.12215039730072022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,power_law_1.2,0.14700160026550294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,power_law_1.2,0.1649664044380188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,power_law_1.2,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,power_law_1.2,0.2592319965362549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,power_law_1.2,0.3372351884841919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,power_law_1.2,0.4130688190460205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,power_law_1.2,0.5833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,power_law_1.2,0.7935423851013184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,balanced,0.04542933404445648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,balanced,0.047482664386431374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,balanced,0.05227733155091604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,balanced,0.05967999994754791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,balanced,0.0804319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,balanced,0.09627733627955119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,balanced,0.11387733618418376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,balanced,0.15614933768908182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,balanced,0.18665599822998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,balanced,0.25335466861724854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,balanced,0.3182613253593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.08766080141067505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.13306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.18545279502868653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.18483200073242187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.19207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.20321919918060302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.20304639339447023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.21539199352264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.2160640001296997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.22124159336090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.2342207908630371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.24238719940185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.24982399940490724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.28216960430145266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.318886399269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.3657792091369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.4083263874053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.5090303897857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.6051519870758056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,0.8565055847167968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.083238410949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,1.5261695861816407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,1.8890687942504882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.09504640102386475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.16811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.21207039356231688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.3518336057662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.38382079601287844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.3986112117767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.40407681465148926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.41800317764282224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.43300480842590333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.4503744125366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.453766393661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.47957119941711424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.2,0.505350399017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.2,0.515283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.2,0.5518015861511231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.2,0.6411136150360107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.2,0.6931327819824219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.2,0.8551872253417969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.0548031806945801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.2,1.3451583862304688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.2,1.6614208221435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.2,2.220358467102051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,power_law_1.01,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.2,2.651379203796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.2,3.7944961547851563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,power_law_1.01,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,power_law_1.01,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.2,4.903238296508789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,power_law_1.01,0.06346240043640136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,power_law_1.01,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,power_law_1.01,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,power_law_1.01,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,power_law_1.01,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,power_law_1.01,0.08743039965629577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,power_law_1.01,0.11191680431365966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.01,0.12714879512786864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.01,0.17296639680862427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.01,0.2061311960220337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.01,0.279040002822876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.01,0.3246592044830322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.01,0.4477695941925049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.01,0.5777599811553955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.01,0.8853631973266601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.01,1.1600064277648925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.08807680010795593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.15317120552062988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.15924479961395263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.15707520246505738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.16117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.16030720472335816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.1623296022415161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.16779520511627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.1705471992492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.1756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1806015968322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.18793599605560302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.20673279762268065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.21222400665283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.22900478839874266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.2678080081939697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3032063961029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.38227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.4609087944030762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.6269887924194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.7656447887420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.1094016075134276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.4356800079345704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,2.140659141540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,2.656844711303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,balanced,0.029669334491093952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,balanced,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,balanced,0.027434666951497395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,balanced,0.036362667878468834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,balanced,0.04270400106906891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,balanced,0.047450666626294456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,balanced,0.060677334666252136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,balanced,0.07252266506354015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,balanced,0.1018933355808258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,balanced,0.12113066514333089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,balanced,0.16686399777730307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,balanced,0.17802667617797852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.09776639938354492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.15021439790725707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.18718080520629882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.21719040870666503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.26940159797668456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.3007424116134644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.2834111928939819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.30938239097595216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.3108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.2834687948226929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.2942975997924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.3023231983184814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.2947711944580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.3059135913848877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.3090111970901489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.3088511943817139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.3197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.35475199222564696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.38937599658966066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.47757439613342284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.4839935779571533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,power_law_1.2,0.5801023960113525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,power_law_1.2,0.613702392578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,power_law_1.2,0.804588794708252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,power_law_1.2,0.8423423767089844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,power_law_1.2,1.2025088310241698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,power_law_1.2,1.5000576019287108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,balanced,0.06708799799283345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,balanced,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,balanced,0.06856533388296764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,balanced,0.06985066831111908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,balanced,0.06871999800205231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,balanced,0.07203733424345653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,balanced,0.07082666456699371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,balanced,0.07618133227030437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,balanced,0.08655466636021932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,balanced,0.0869760016600291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,balanced,0.08821333448092143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,balanced,0.10204799969991048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.10301867127418518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.11718933780988057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.1397760013739268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.1755573352177938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.19642666975657144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.33108800649642944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.4678293466567993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,balanced,0.6017813285191854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,power_law_1.01,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,power_law_1.01,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,power_law_1.01,0.19998719692230224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,power_law_1.01,0.20703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,power_law_1.01,0.20613760948181153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,power_law_1.01,0.20351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,power_law_1.01,0.21472640037536622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,power_law_1.01,0.21726078987121583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,power_law_1.01,0.21981439590454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,power_law_1.01,0.22188799381256102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,power_law_1.01,0.2279423952102661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,power_law_1.01,0.23074560165405272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,power_law_1.01,0.25096960067749025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,power_law_1.01,0.2748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,power_law_1.01,0.2927743911743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,power_law_1.01,0.2844928026199341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,power_law_1.01,0.32737278938293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,power_law_1.01,0.37356159687042234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,power_law_1.01,0.456166410446167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,power_law_1.01,0.5316927909851075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,power_law_1.01,0.7130047798156738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,power_law_1.01,0.8889151573181152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,power_law_1.01,1.2292160034179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,power_law_1.01,1.5966591835021973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,power_law_1.01,2.328236770629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,power_law_1.01,3.0591039657592773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.14999680519104003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.24165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.3134592056274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.3874111890792847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.4628928184509277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.5217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.6968063831329345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.727513599395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.7137728214263916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.7383679866790771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.7754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,power_law_1.01,0.701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.7019904136657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,power_law_1.01,0.7101183891296386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,power_law_1.01,0.7712704181671143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,power_law_1.01,0.7628223896026611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,power_law_1.01,0.8120063781738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,power_law_1.01,0.8515968322753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,power_law_1.01,0.8808064460754395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,power_law_1.01,1.0543295860290527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,power_law_1.01,0.9878144264221191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,power_law_1.01,1.1225600242614746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,power_law_1.01,1.2627200126647948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,power_law_1.01,1.6136064529418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,power_law_1.01,1.734752082824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,power_law_1.01,2.2500799179077147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,power_law_1.01,2.945318412780762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.04782080054283142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.06771199703216553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.12792960405349732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.1522752046585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.19813120365142822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.24049921035766603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.3232896089553833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.4844799995422363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.04996480047702789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.10224640369415283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.21156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.2607232093811035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.3393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.5323200225830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,power_law_1.01,0.6428864002227783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,balanced,0.04422933359940847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,balanced,0.04980800052483877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,balanced,0.051498666405677795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,balanced,0.05778666834036509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,balanced,0.06234133243560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,balanced,0.09050666292508443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,balanced,0.11156266927719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,balanced,0.14443733294804892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,balanced,0.20361600319544473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,balanced,0.24160534143447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,balanced,0.33718399206797284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,balanced,0.43403200308481854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,power_law_1.2,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,power_law_1.2,0.06613119840621948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,power_law_1.2,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,power_law_1.2,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,power_law_1.2,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,power_law_1.2,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,power_law_1.2,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,power_law_1.2,0.09582719802856446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.2,0.1245568037033081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.2,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.2,0.18688640594482422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.2,0.2684416055679321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.2,0.30976641178131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.2,0.48844161033630373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.2,0.7481088161468505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.2,0.9700736045837403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.10861439704895019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.12567039728164672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.2,1.3966527938842774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.180511999130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.2142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.2989887952804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.3888256072998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,balanced,0.06638399759928386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,balanced,0.08002133170763652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,balanced,0.07981866598129272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,balanced,0.08085866769154866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,balanced,0.08179733157157898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,balanced,0.08145066599051158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,balanced,0.08142399787902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,balanced,0.08517866333325703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,balanced,0.0883146623770396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,balanced,0.08715732892354329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,balanced,0.0865760048230489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,balanced,0.09226666887601216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,balanced,0.10048533479372661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,balanced,0.10566932956377666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,balanced,0.11518399914105733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,balanced,0.15495466192563376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,balanced,0.1576746702194214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,balanced,0.2172586719195048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,balanced,0.2445546587308248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,balanced,0.34253867467244464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,balanced,0.43270933628082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,balanced,0.6373813152313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,balanced,0.7932106653849283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,balanced,0.09416533509890239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,balanced,0.13607999682426453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,balanced,0.14071999986966452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,balanced,0.14054399728775024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,balanced,0.14164800445238748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,balanced,0.14316800236701965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,balanced,0.14379200339317322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,balanced,0.14854400356610617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,balanced,0.15105066696802774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,balanced,0.15205333630243936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,balanced,0.16033066312472025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,balanced,0.16527466972668967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,balanced,0.17227200667063394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,balanced,0.18547733624776205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,balanced,0.2185493310292562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.24010666211446127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.29890666405359906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.34913599491119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.5060106515884399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.6105546553929647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.8864746888478597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,balanced,1.1303306420644124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,balanced,1.66538667678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,balanced,2.183882713317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.09385600090026855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.13727359771728515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.15845760107040405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.22400639057159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.2720896005630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.3962687969207764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,power_law_1.2,0.5111104011535644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,power_law_1.2,0.770521593093872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,power_law_1.2,1.1055744171142579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,balanced,0.08084799846013387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,balanced,0.07474666833877563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,balanced,0.07189333438873291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,balanced,0.07227733234564464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,balanced,0.07291199763615926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,balanced,0.07435200115044911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,balanced,0.07352533439795177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,balanced,0.07627200086911519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,balanced,0.07662400106589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,balanced,0.07950399816036224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,balanced,0.08500799536705017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,balanced,0.08884800473848979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,balanced,0.09584533174832661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.10511466860771179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.11782933274904887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.1357439955075582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.16955200831095377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.20548266172409058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.26518932978312176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.33294934034347534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.46856534481048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,balanced,0.6189013322194418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.07551360130310059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.1
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.11607680320739747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.20745599269866943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.2678015947341919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.3324863910675049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.4305920124053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.6095168113708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,power_law_1.2,0.7976831912994384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.0851967811584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.01,0.07445120215415954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.01,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.01,0.06917120218276977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.01,0.07089920043945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.01,0.07112320065498352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.01,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.01,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.01,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.01,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.01,0.11181440353393554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.01,0.12581119537353516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.01,0.16000640392303467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.01,0.18584959506988524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.01,0.24085121154785155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.01,0.2506943941116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.01,0.31912961006164553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.01,0.4028736114501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.01,0.5301119804382324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.01,0.6691840171813965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,power_law_1.01,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,power_law_1.01,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,power_law_1.01,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,power_law_1.01,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,power_law_1.01,0.06940799951553345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,power_law_1.01,0.0937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,power_law_1.01,0.11889280080795288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,power_law_1.01,0.16563199758529662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,power_law_1.01,0.22278399467468263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,power_law_1.01,0.28517119884490966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,power_law_1.01,0.37118079662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.09446399807929992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.09249280095100403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.09089279770851136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.09697920083999634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.11005439758300781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.13027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.15463680028915405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.170687997341156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.204915189743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.2483135938644409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.3205696105957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.4029888153076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,power_law_1.01,0.5532224178314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,power_law_1.01,0.6809216022491456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,power_law_1.01,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,power_law_1.01,0.07737600207328796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,power_law_1.01,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,power_law_1.01,0.10893440246582031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,power_law_1.01,0.13523839712142943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,power_law_1.01,0.15555839538574218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,power_law_1.01,0.18138240575790404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,power_law_1.01,0.19205119609832763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,power_law_1.01,0.18925440311431885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,power_law_1.01,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,power_law_1.01,0.2097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,power_law_1.01,0.2145087957382202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,power_law_1.01,0.22184319496154786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,power_law_1.01,0.2361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,power_law_1.01,0.24268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,power_law_1.01,0.2667520046234131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,power_law_1.01,0.307206392288208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,power_law_1.01,0.32328319549560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,power_law_1.01,0.3855743885040283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,power_law_1.01,0.44595842361450194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,power_law_1.01,0.5739391803741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,power_law_1.01,0.6819712162017822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,power_law_1.01,0.9386688232421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,power_law_1.01,1.1748096466064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,power_law_1.01,1.6305984497070312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,power_law_1.01,2.042073631286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,power_law_1.01,0.059520000219345094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,power_law_1.01,0.07107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,power_law_1.01,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,power_law_1.01,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,power_law_1.01,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,power_law_1.01,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,power_law_1.01,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,power_law_1.01,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,power_law_1.01,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,power_law_1.01,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,power_law_1.01,0.10028159618377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,power_law_1.01,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,power_law_1.01,0.1272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,power_law_1.01,0.1569983959197998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,power_law_1.01,0.1758911967277527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,power_law_1.01,0.2288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,power_law_1.01,0.27212159633636473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,power_law_1.01,0.36910080909729004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,power_law_1.01,0.48293118476867675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,power_law_1.01,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,power_law_1.01,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,power_law_1.01,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,power_law_1.01,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,power_law_1.01,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,power_law_1.01,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,power_law_1.01,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,power_law_1.01,0.11255040168762206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,power_law_1.01,0.13434879779815673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,power_law_1.01,0.1459712028503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,power_law_1.01,0.19396480321884155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,power_law_1.01,0.24512639045715331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,power_law_1.01,0.3177151918411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,power_law_1.01,0.4344639778137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,power_law_1.01,0.6353919982910157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,power_law_1.01,0.7694528102874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,power_law_1.01,1.1364095687866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,power_law_1.01,1.709939193725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.01,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.01,0.029145601391792297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.01,0.02995840013027191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.01,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.01,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.01,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.01,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.01,0.06814079880714416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.01,0.12136960029602051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.01,0.17191679477691652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.01,0.21475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,power_law_1.2,0.07452800273895263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,power_law_1.2,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,power_law_1.2,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,power_law_1.2,0.07626879811286927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,power_law_1.2,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,power_law_1.2,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,power_law_1.2,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,power_law_1.2,0.100108802318573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,power_law_1.2,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,power_law_1.2,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.2,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.2,0.15005439519882202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.2,0.17384320497512817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.2,0.2219327926635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.2,0.27932798862457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.2,0.39052159786224366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.2,0.5211775779724122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.2,0.7340352058410644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.2,1.0189184188842773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,power_law_1.01,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,power_law_1.01,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,power_law_1.01,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,power_law_1.01,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,power_law_1.01,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,power_law_1.01,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,power_law_1.01,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,power_law_1.01,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,power_law_1.01,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,power_law_1.01,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,power_law_1.01,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.01,0.08012800216674805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.01,0.09456639885902404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.01,0.11167360544204712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.01,0.14252159595489503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.01,0.17276159524917603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.01,0.2256256103515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.01,0.2694144010543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.01,0.40025601387023924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.01,0.4791935920715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,balanced,0.052895997961362205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,balanced,0.0543039987484614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,balanced,0.054789334535598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,balanced,0.05779733260472616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,balanced,0.0606879989306132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,balanced,0.06489600241184235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,balanced,0.07390933235486348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,balanced,0.0832533339659373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,balanced,0.0986346701780955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,balanced,0.11681066950162251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,balanced,0.13702932993570963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,balanced,0.16114667057991028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,balanced,0.2243573268254598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,balanced,0.2813813289006551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,balanced,0.4074133237202962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,balanced,0.5157333215077718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,balanced,0.7560586929321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,balanced,0.9885866641998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.21367039680480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.3413248062133789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4445312023162842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.7101119995117188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.6833792209625245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.7176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.7378240108489991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.748089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.7651519775390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.7770112037658692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,power_law_1.2,0.7930111885070801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,power_law_1.2,0.8220607757568359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,power_law_1.2,0.8328639984130859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,power_law_1.2,0.8304575920104981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,power_law_1.2,0.8781760215759278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,power_law_1.2,0.9072256088256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,power_law_1.2,0.9658304214477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,power_law_1.2,1.1198399543762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,power_law_1.2,1.2805503845214843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,power_law_1.2,1.5142911911010741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,power_law_1.2,1.7309696197509765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,power_law_1.2,2.2555456161499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,power_law_1.2,2.787379264831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,power_law_1.2,3.9400447845458983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,power_law_1.2,5.115283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,balanced,0.05045866469542185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,balanced,0.07006933291753133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,balanced,0.1011946698029836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,balanced,0.15500799814860025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,balanced,0.2537546753883362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,balanced,0.2555946707725525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,balanced,0.2619306643803914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,balanced,0.26073066393534344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,balanced,0.26422399282455444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,balanced,0.2654986580212911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,balanced,0.26976533730824787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,balanced,0.2709386746088664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,balanced,0.28173865874608356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,balanced,0.2900000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,balanced,0.2935840090115865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,balanced,0.3083146611849467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,balanced,0.3324586749076843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,balanced,0.3514346679051717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,balanced,0.39870933691660565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,balanced,0.4448053439458211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,balanced,0.5664373238881429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,balanced,0.6585706472396851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,balanced,0.8961226940155029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,balanced,1.0520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,balanced,1.527546723683675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,balanced,1.921237309773763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,power_law_1.2,7.649126434326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,balanced,0.02773866554101308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,balanced,0.02939733366171519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,balanced,0.02920000006755193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,balanced,0.029290666182835896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,balanced,0.032602667808532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,balanced,0.04608533283074697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,balanced,0.06039999922116598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,balanced,0.0746613343556722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,balanced,0.09875733653704326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,balanced,0.12455999851226807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,balanced,0.14890666802724203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,balanced,0.2060426672299703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,balanced,0.31487466891606647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,power_law_1.2,9.991641235351562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,balanced,0.05458133419354757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,balanced,0.057402665416399636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,balanced,0.0566293348868688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,balanced,0.066021333138148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,balanced,0.07875200112660725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,balanced,0.10146133104960124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,balanced,0.10412266850471497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,balanced,0.13380799690882364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,balanced,0.15309866269429526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,balanced,0.20452266931533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,balanced,0.24442134300867716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,balanced,0.3355413277943929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,balanced,0.4193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07089920043945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,power_law_1.01,0.11612800359725953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,power_law_1.01,0.17315200567245484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,power_law_1.01,0.17900160551071168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,power_law_1.01,0.1881600022315979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,power_law_1.01,0.18878079652786256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,power_law_1.01,0.19386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,power_law_1.01,0.1996608018875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,power_law_1.01,0.20210559368133546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,power_law_1.01,0.20234239101409912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,power_law_1.01,0.20885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,power_law_1.01,0.21724159717559816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,power_law_1.01,0.22288639545440675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,power_law_1.01,0.22776319980621337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,power_law_1.01,0.2581824064254761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.2696768045425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.31699199676513673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.34250240325927733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.4242432117462158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.5208896160125732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,power_law_1.01,0.6732416152954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,power_law_1.01,0.8227007865905762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,power_law_1.01,1.123628807067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,power_law_1.01,1.4252096176147462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,balanced,0.09913067022959392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,balanced,0.14851199587186178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,balanced,0.24596265951792398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,balanced,0.439301331837972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,balanced,0.8362027009328207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,balanced,1.228122631708781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,balanced,1.2200533548990886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,balanced,1.225215991338094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,balanced,1.2277973492940266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,balanced,1.2311466534932454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,balanced,1.2389973004659016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,balanced,1.2431573073069255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,balanced,1.252778689066569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,balanced,1.2642079989115398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,balanced,1.2693333625793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,balanced,1.285973310470581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,balanced,1.3163093725840251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,balanced,1.3437387148539226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,balanced,1.404736042022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,balanced,1.4641493161519368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,balanced,1.5913920402526855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,balanced,1.6875839233398438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,balanced,1.8708853721618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,balanced,2.1670079231262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,balanced,2.541744073232015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,balanced,3.3417491912841797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,power_law_1.01,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,power_law_1.01,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,power_law_1.01,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,power_law_1.01,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,power_law_1.01,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,power_law_1.01,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,power_law_1.01,0.08208640217781067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,power_law_1.01,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,power_law_1.01,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,power_law_1.01,0.09122560024261475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,power_law_1.01,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,power_law_1.01,0.0979968011379242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,power_law_1.01,0.10270719528198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,power_law_1.01,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,power_law_1.01,0.1317055940628052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,power_law_1.01,0.15946240425109864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,power_law_1.01,0.16865919828414916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,power_law_1.01,0.23594880104064941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,power_law_1.01,0.269651198387146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,power_law_1.01,0.38577919006347655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,power_law_1.01,0.448748779296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,balanced,0.04251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,balanced,0.051455999414126076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,balanced,0.05379733443260193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,balanced,0.0662773350874583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,balanced,0.06647466619809468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,balanced,0.08692800005276997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,balanced,0.09714133540789287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,balanced,0.12340799967447917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,balanced,0.13658666610717773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,balanced,0.18283732732137045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,balanced,0.2074399987856547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,balanced,0.2863893310228984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,balanced,0.34931735197703045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.04785920083522797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.06510080099105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.08670719861984252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.10895359516143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.14841599464416505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.17125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.22247040271759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.2721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.3772671937942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.4777088165283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.6782656192779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.878809642791748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,power_law_1.2,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,power_law_1.2,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,power_law_1.2,0.06728960275650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,power_law_1.2,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,power_law_1.2,0.08766080141067505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,power_law_1.2,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,power_law_1.2,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,power_law_1.2,0.09151359796524047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,power_law_1.2,0.09034879803657532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,power_law_1.2,0.09782400131225585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,power_law_1.2,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,power_law_1.2,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,power_law_1.2,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,power_law_1.2,0.10613119602203369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,power_law_1.2,0.11319040060043335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,power_law_1.2,0.12280319929122925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.2,0.13655680418014526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.2,0.16132479906082153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.2,0.17950719594955444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.2,0.2219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.2,0.2704511880874634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.2,0.36241281032562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.2,0.4368127822875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.2,0.5825984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.2,0.7295423984527588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.07671040296554565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.10654720067977905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.13274240493774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.15438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.20010879039764404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,balanced,0.08894399801890056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,balanced,0.12548800309499106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,balanced,0.20177600781122842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.23462400436401368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,balanced,0.3468639850616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,balanced,0.608730673789978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.2984447956085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,balanced,0.8338613510131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.38607358932495117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,balanced,0.839792013168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,balanced,0.8408053716023763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.553875207901001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,balanced,0.8438133398691813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,balanced,0.8463253180185953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.6984384059906006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,balanced,0.8516320387522379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,balanced,0.85371200243632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,balanced,0.8561813036600748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,balanced,0.8616639773050944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,balanced,0.8722933133443197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,balanced,0.877013365427653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,balanced,0.8887253602345785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,balanced,1.048624038696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,balanced,0.9420586427052816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,balanced,1.1419466336568196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,balanced,1.0474666754404705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,balanced,1.5092906951904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,balanced,1.235317309697469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,balanced,2.0124425888061523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,balanced,1.8274879455566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,balanced,3.132330576578776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,balanced,2.792517344156901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,power_law_1.01,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,power_law_1.01,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,power_law_1.01,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,power_law_1.01,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,power_law_1.01,0.02975359857082367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,power_law_1.01,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,power_law_1.01,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,power_law_1.01,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.07650560140609741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.07809919714927674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.09476479887962341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.10922240018844605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.12583039999008178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.15594880580902098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,power_law_1.01,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.2224191904067993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.2587712049484253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.3588543891906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.4319615840911865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.6269696235656739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.8469632148742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,power_law_1.01,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,power_law_1.01,0.13409279584884642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,power_law_1.01,0.19022079706192016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,power_law_1.01,0.231276798248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.12401920557022095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.15913599729537964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.20842878818511962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.25744640827178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.34542078971862794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.34101760387420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.3386751890182495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.36355841159820557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.381062388420105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.3942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.3891711950302124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.41434240341186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.42409601211547854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.440883207321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.463756799697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.5258304119110108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.5337664127349854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.6235392093658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.6579648017883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,power_law_1.01,0.3390144109725952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,power_law_1.2,0.7874752044677734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,power_law_1.2,0.9082176208496093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,power_law_1.2,1.1382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,power_law_1.2,1.390681552886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,power_law_1.2,1.8505535125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,power_law_1.2,2.390630340576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,power_law_1.01,0.4630080223083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,balanced,0.047397335370381675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,balanced,0.047397335370381675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,balanced,0.054330666859944664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,balanced,0.05821866790453593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,balanced,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,balanced,0.061679999033610024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,balanced,0.06567466755708058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,balanced,0.06715199848016103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,balanced,0.07031466563542683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,balanced,0.06961066524187724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,balanced,0.08406399687131245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,balanced,0.08917867143948872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,balanced,0.10594666997591655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,balanced,0.11596799890200298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,balanced,0.13169599572817484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,balanced,0.16564266880353293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,balanced,0.19088532527287802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,balanced,0.2600853244463603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,balanced,0.32236266136169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,balanced,0.46519466241200763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,balanced,0.604853351910909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,power_law_1.01,0.05679360032081604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,power_law_1.01,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,power_law_1.01,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,power_law_1.01,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,power_law_1.01,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,power_law_1.01,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,power_law_1.01,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,power_law_1.01,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.01,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.01,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.01,0.09660800099372864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.01,0.14269440174102782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.01,0.19760639667510987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.01,0.23918719291687013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.01,0.3313407897949219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.01,0.4272831916809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,power_law_1.2,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,power_law_1.2,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,power_law_1.2,0.08026880025863647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,power_law_1.2,0.12024320363998413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.2,0.13575040102005004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.2,0.16416000127792357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.2,0.20468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.2,0.3087296009063721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.2,0.4469696044921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.2,0.6132927894592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.2,0.8041855812072753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,balanced,0.09485866626103719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,balanced,0.14405333002408346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,balanced,0.24238399664560953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,balanced,0.439797321955363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,balanced,0.836512009302775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,balanced,1.6176479657491047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,balanced,2.403893311818441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,balanced,2.4107093811035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,balanced,2.4182666142781577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,balanced,2.4232053756713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,balanced,2.4295093218485513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,balanced,2.4382400512695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,balanced,2.4517173767089844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,balanced,2.4547573725382485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,balanced,2.4708213806152344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,balanced,2.478384017944336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,balanced,2.505557378133138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,balanced,2.564250628153483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,balanced,2.6081973711649575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,balanced,2.7029012044270835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,balanced,2.804885228474935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,balanced,2.999802589416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,balanced,3.178229331970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,balanced,3.582634607950846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,balanced,4.088234583536784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,balanced,4.663093249003093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,balanced,6.042687733968099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.2,1.2299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,power_law_1.01,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,power_law_1.01,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,power_law_1.01,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,power_law_1.01,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,power_law_1.01,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,power_law_1.01,0.06320639848709106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,power_law_1.01,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,power_law_1.01,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,power_law_1.01,0.1002303957939148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,power_law_1.01,0.1159551978111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,power_law_1.01,0.14881919622421264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,power_law_1.01,0.1752832055091858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,power_law_1.01,0.2348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,power_law_1.01,0.29024639129638674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.2,1.4336511611938476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,power_law_1.01,0.41246719360351564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,power_law_1.01,0.5574336051940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,power_law_1.01,0.7366784095764161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,power_law_1.01,1.0364352226257325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.01,0.09157119989395142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.01,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.01,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.01,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.01,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.01,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.01,0.09310719966888428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.01,0.0942143976688385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.01,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.01,0.09557120203971863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.01,0.0980351984500885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.01,0.10051200389862061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.01,0.1102720022201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.01,0.11159679889678956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.01,0.12100479602813721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.01,0.13946880102157594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.01,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.01,0.196070396900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.01,0.22123520374298095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.01,0.28580479621887206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.01,0.3076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.01,0.4111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.01,0.5093632221221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.01,0.761900806427002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.01,0.9379008293151856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,power_law_1.01,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,power_law_1.01,0.09502720236778259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,power_law_1.01,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,power_law_1.01,0.12778879404067994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,power_law_1.01,0.12120319604873657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,power_law_1.01,0.12905600070953369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,power_law_1.01,0.12503039836883545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,power_law_1.01,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,power_law_1.01,0.12608640193939208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,power_law_1.01,0.1298624038696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,power_law_1.01,0.12961920499801635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,power_law_1.01,0.1325824022293091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,power_law_1.01,0.1322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,power_law_1.01,0.13772159814834595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,power_law_1.01,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,power_law_1.01,0.14588160514831544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,power_law_1.01,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.01,0.16543359756469728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,balanced,0.044453332821528115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,balanced,0.047024001677831016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.01,0.18856960535049438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,balanced,0.0462773342927297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.01,0.21210880279541017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,balanced,0.04906133313973745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.01,0.2612544059753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,balanced,0.06560533245404561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,balanced,0.06766400237878163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,balanced,0.08408000071843465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.01,0.3042304039001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,balanced,0.09086400270462036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,balanced,0.11002666751543681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,balanced,0.13371733824412027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.01,0.41992959976196287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,balanced,0.1832373340924581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,balanced,0.2433919906616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,balanced,0.3489866654078166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.01,0.4985856056213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,balanced,0.4546560049057007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,balanced,0.6564480066299438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.01,0.7261568069458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,balanced,0.8538933595021566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.01,0.9982975959777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.01,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.01,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.01,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.01,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.01,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.01,0.1528831958770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.01,0.1895616054534912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.01,0.2658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.01,0.3194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.08222079873085023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.10565760135650634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.13417600393295287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.1791103959083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.1824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.18484480381011964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.18913279771804808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.19404799938201905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.20071039199829102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.20463359355926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.21269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.21948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.22767999172210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.23135359287261964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.25338239669799806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.2700416088104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.3220736026763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.3381632089614868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.4241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,power_law_1.01,0.4824384212493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,power_law_1.01,0.6227136135101319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,power_law_1.01,0.8176704406738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,power_law_1.01,1.084780788421631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,power_law_1.01,1.3782527923583985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.01,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.01,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.01,0.0606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.01,0.05842559933662415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.01,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.01,0.0611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.01,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.01,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.01,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.01,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.01,0.10835200548171997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.01,0.13150080442428588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.01,0.1560639977455139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.01,0.20395519733428955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.01,0.20992639064788818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.01,0.2704511880874634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.01,0.3380800008773804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.01,0.46449918746948243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.01,0.5817279815673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,1,power_law_1.2,0.1525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,1,power_law_1.2,0.1753983974456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,1,power_law_1.2,0.20621440410614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,1,power_law_1.2,0.2598400115966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,1,power_law_1.2,0.32235519886016845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,1,power_law_1.2,0.401855993270874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,1,power_law_1.2,0.5415872097015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,1,power_law_1.2,0.5641280174255371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,1,power_law_1.2,0.591161584854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,1,power_law_1.2,0.5879360198974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,1,power_law_1.2,0.6240960121154785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,1,power_law_1.2,0.6417664051055908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,1,power_law_1.2,0.6594367980957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,1,power_law_1.2,0.6731264114379882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,1,power_law_1.2,0.6846464157104493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,1,power_law_1.2,0.6917888164520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,1,power_law_1.2,0.7247424125671387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,1,power_law_1.2,0.7713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.2,0.808249568939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.2,0.8854720115661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.2,0.9882880210876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.2,1.1393343925476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.2,1.3293567657470704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.2,1.7372991561889648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.2,2.121913528442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.2,2.9294208526611327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.2,3.7744831085205077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,balanced,0.06272533535957336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,balanced,0.048138668139775596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,balanced,0.05225066840648651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,balanced,0.053727999329566956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,balanced,0.06308266520500183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,balanced,0.06311466793219249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,balanced,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.09483733773231506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.10944533348083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.13788800438245138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.16447466611862183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.213754673798879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.2572213411331177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.3555413484573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.45281068483988446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.08255360126495362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.09629439711570739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.1204416036605835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.1412608027458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.17730560302734374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.23418240547180175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.3150719881057739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.3458303928375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,power_law_1.2,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,power_law_1.2,0.07681279778480529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,power_law_1.2,0.07727360129356384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,power_law_1.2,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,power_law_1.2,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,power_law_1.2,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,power_law_1.2,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,power_law_1.2,0.0910975992679596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,power_law_1.2,0.09409919977188111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,power_law_1.2,0.10031360387802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,power_law_1.2,0.11234560012817382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.2,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.2,0.1510655999183655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.2,0.17600640058517455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.2,0.2246527910232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.2,0.26080000400543213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.2,0.348307204246521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.2,0.469817590713501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.2,0.6943039894104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.2,0.8612095832824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,power_law_1.2,0.557689619064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,power_law_1.2,0.6982912063598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,power_law_1.2,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,power_law_1.2,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,power_law_1.2,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,power_law_1.2,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,power_law_1.2,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,power_law_1.2,0.12179839611053467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,power_law_1.2,0.12259199619293212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,power_law_1.2,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,power_law_1.2,0.13454079627990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,power_law_1.2,0.13516160249710082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,power_law_1.2,0.1380288004875183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,power_law_1.2,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.1684991955757141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.20097920894622803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.2135103940963745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.2632767915725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.3174911975860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.38753280639648435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,power_law_1.2,0.4763775825500488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,power_law_1.2,0.6380288124084472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,power_law_1.2,0.870201587677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,power_law_1.01,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,power_law_1.01,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,power_law_1.01,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,power_law_1.01,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,power_law_1.01,0.04008319973945618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,power_law_1.01,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,power_law_1.01,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,power_law_1.01,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,power_law_1.01,0.11370240449905396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,power_law_1.01,0.15754879713058473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,power_law_1.01,0.19715839624404907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.15014400482177734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.1972991943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.27088000774383547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.45024638175964354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.47251200675964355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5060991764068603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5098624229431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.537011194229126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.5489344120025634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.5667776107788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.5707071781158447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.5826047897338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.593120002746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.620684814453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.6409215927124023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.6925631999969483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.7379519939422607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,power_law_1.2,0.8888575553894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,power_law_1.2,0.9753215789794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.2057215690612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,power_law_1.2,1.267251205444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,power_law_1.2,1.6218687057495118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,power_law_1.2,2.243449592590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,power_law_1.2,2.919968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,power_law_1.2,3.3912128448486327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,balanced,0.04458666841189066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,balanced,0.047557334105173744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,balanced,0.04950400193532308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,balanced,0.05239466826121012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,balanced,0.05629333357016245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,balanced,0.06399466594060262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,balanced,0.07438933352629344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,balanced,0.08201066652933757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,balanced,0.10779199997584026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,balanced,0.1257866621017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,balanced,0.1666826605796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,balanced,0.18853867053985596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,balanced,0.2593653400739034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,balanced,0.30315200487772626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,balanced,0.4333866834640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,balanced,0.5481333335240682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,power_law_1.2,0.05139840245246887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,power_law_1.2,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,power_law_1.2,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,power_law_1.2,0.0692736029624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,power_law_1.2,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,power_law_1.2,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,power_law_1.2,0.05934079885482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,power_law_1.2,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,power_law_1.2,0.08880640268325805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,power_law_1.2,0.09225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,power_law_1.2,0.06709759831428527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,power_law_1.2,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,power_law_1.2,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,power_law_1.2,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,power_law_1.2,0.09735040068626404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,power_law_1.2,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,power_law_1.2,0.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,power_law_1.2,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,power_law_1.2,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,power_law_1.2,0.11662720441818238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,power_law_1.2,0.10485119819641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,power_law_1.2,0.1187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,power_law_1.2,0.12718720436096193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,power_law_1.2,0.14695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,power_law_1.2,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,power_law_1.2,0.17319680452346803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,power_law_1.2,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,power_law_1.2,0.15996160507202148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,power_law_1.2,0.20477440357208251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,power_law_1.2,0.2760512113571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,power_law_1.2,0.191212797164917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,power_law_1.2,0.26835200786590574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,power_law_1.2,0.3568639993667603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,power_law_1.2,0.4162816047668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,power_law_1.2,0.3601984024047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,power_law_1.2,0.5571584224700927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,power_law_1.2,0.4826176166534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,power_law_1.2,0.7566847801208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,power_law_1.2,0.6792448043823243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,power_law_1.2,0.8666815757751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,power_law_1.2,1.166975975036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,power_law_1.2,1.3914048194885253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,balanced,0.035088000198205314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,balanced,0.03711466739575068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,balanced,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,balanced,0.0383093332250913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,balanced,0.10930132865905762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,balanced,0.14689067006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,balanced,0.1769333283106486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,power_law_1.2,1.1538496017456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.08250240087509156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.08504319787025452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.08638079762458802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.0862335979938507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10147839784622192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.1265663981437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.14813439846038817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.16172800064086915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.19569920301437377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.24396159648895263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.3190592050552368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.41326079368591306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.5455808162689209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.7262720108032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,power_law_1.01,0.028384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,power_law_1.01,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,power_law_1.01,0.028313601016998292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,power_law_1.01,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,power_law_1.01,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,power_law_1.01,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,power_law_1.01,0.08514559864997864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,power_law_1.01,0.11585919857025147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.10569599866867066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.1271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.1508288025856018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,power_law_1.01,0.15315200090408326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.1573184013366699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,power_law_1.01,0.17424000501632692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.19891200065612794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.19571199417114257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.19867520332336425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.20831360816955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,power_law_1.01,0.21273601055145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.2137984037399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.22016639709472657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.22775039672851563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.23697280883789062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.25021440982818605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.25322880744934084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.2748480081558228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.3161535978317261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.36784000396728517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.4108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.527180814743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.6436223983764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,0.8348416328430176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.121132755279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,1.510316753387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,2.008140754699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,power_law_1.01,0.3187135934829712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,power_law_1.01,0.4139840126037598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,power_law_1.01,0.5785727977752686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.01,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.01,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.01,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.01,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.01,0.06714239716529846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.01,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.01,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.01,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.01,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.01,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.01,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.01,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.01,0.11847679615020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.01,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.01,0.16858240365982055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.01,0.21747200489044188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.01,0.2269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.01,0.2981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.01,0.3408384084701538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.01,0.4960768222808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.01,0.6334208011627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,balanced,0.12013866504033406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,balanced,0.20349333683649698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,balanced,0.3698933521906535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,balanced,0.7040586471557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,balanced,0.7045333385467529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,balanced,0.7083893616994222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,balanced,0.7114346822102865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,balanced,0.7151040236155192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,balanced,0.7154186566670736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,balanced,0.7194506327311198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,balanced,0.7266879876454672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,balanced,0.7295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,balanced,0.7371306419372559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,balanced,0.743834654490153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,balanced,0.7507733503977457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,balanced,0.7701013088226318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,balanced,0.7976160049438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,balanced,0.8337066968282064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,balanced,0.8786826928456625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,balanced,0.9210933049519857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,balanced,1.1310773690541585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,balanced,1.151136000951131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,balanced,1.5724587440490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,balanced,1.7156000137329102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,balanced,2.3842453956604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,balanced,3.0907198588053384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,power_law_1.01,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,power_law_1.01,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,power_law_1.01,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,power_law_1.01,0.12849279642105102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,power_law_1.01,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,power_law_1.01,0.13086719512939454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,power_law_1.01,0.12392319440841675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,power_law_1.01,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,power_law_1.01,0.12472959756851196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,power_law_1.01,0.12769919633865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,power_law_1.01,0.12966400384902954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,power_law_1.01,0.14176640510559083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,power_law_1.01,0.14120960235595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,power_law_1.01,0.15217920541763305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,power_law_1.01,0.15575679540634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.01,0.1751360058784485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.01,0.19844479560852052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.01,0.23739519119262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.01,0.30829439163208006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.01,0.3721024036407471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.01,0.5192575931549073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.01,0.6568575859069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.01,0.9280256271362305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.01,1.2125311851501466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,power_law_1.2,0.030073601007461547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,power_law_1.2,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,power_law_1.2,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,power_law_1.2,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,power_law_1.2,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,power_law_1.2,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,power_law_1.2,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,power_law_1.2,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,power_law_1.2,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,power_law_1.2,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,power_law_1.2,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,power_law_1.2,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,power_law_1.2,0.1721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,power_law_1.2,0.24009599685668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,power_law_1.2,0.3141119956970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,power_law_1.2,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,power_law_1.2,0.04327679872512817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,power_law_1.2,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,power_law_1.2,0.1117311954498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,power_law_1.2,0.13519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,power_law_1.2,0.20292479991912843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,power_law_1.2,0.21873281002044678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,power_law_1.2,0.308025598526001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,power_law_1.2,0.4205632209777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,power_law_1.01,0.09695360064506531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,power_law_1.01,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,power_law_1.01,0.12911360263824462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,power_law_1.01,0.1343808054924011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,power_law_1.01,0.139193594455719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,power_law_1.01,0.137881600856781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,power_law_1.01,0.1422592043876648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,power_law_1.01,0.1388543963432312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,power_law_1.01,0.14039679765701293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,power_law_1.01,0.15013760328292847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,power_law_1.01,0.1536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,power_law_1.01,0.1673408031463623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,power_law_1.01,0.17968640327453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,power_law_1.01,0.18350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,power_law_1.01,0.19648000001907348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,power_law_1.01,0.22437760829925538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,power_law_1.01,0.2558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,power_law_1.01,0.30243840217590334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,power_law_1.01,0.3623039960861206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,power_law_1.01,0.48251519203186033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,power_law_1.01,0.5938111782073975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,power_law_1.01,0.8235327720642089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,power_law_1.01,1.059878444671631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,power_law_1.01,1.5210240364074707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,balanced,0.08010666569073994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,balanced,0.08116266628106435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,balanced,0.08319466809431712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,balanced,0.08460799853006999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,balanced,0.08710933725039165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,balanced,0.10308266679445903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,balanced,0.10222933689753215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,balanced,0.10302399595578511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,balanced,0.11964266498883565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,balanced,0.12130666772524516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,balanced,0.15109333395957947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,balanced,0.16530666748682657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,balanced,0.2230559984842936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,balanced,0.2663946747779846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,balanced,0.370469331741333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,balanced,0.47035733858744305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,balanced,0.6744960149129232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,balanced,0.8720266819000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,power_law_1.01,1.9895488739013671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,power_law_1.01,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,power_law_1.01,0.10253440141677857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,power_law_1.01,0.12780159711837769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,power_law_1.01,0.1798848032951355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,power_law_1.01,0.19315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,power_law_1.01,0.19664000272750853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,power_law_1.01,0.20051839351654052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,power_law_1.01,0.2056960105895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,power_law_1.01,0.20765440464019774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,power_law_1.01,0.22043519020080565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,power_law_1.01,0.22272639274597167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,power_law_1.01,0.23235840797424318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,power_law_1.01,0.24558720588684083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,power_law_1.01,0.2532671928405762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,power_law_1.01,0.2651072025299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,power_law_1.01,0.2961983919143677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,power_law_1.01,0.3274048089981079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,power_law_1.01,0.40427517890930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,power_law_1.01,0.4699264049530029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,power_law_1.01,0.609772777557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,power_law_1.01,0.7418047904968261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,power_law_1.01,1.0072192192077636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,power_law_1.01,1.28220796585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,power_law_1.01,1.8189632415771484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,power_law_1.01,2.2646015167236326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.08695679903030396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.0887615978717804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.09316480159759521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.10423680543899536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.10871679782867431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.11480319499969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.1269695997238159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.14555519819259644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.17184640169143678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.19071999788284302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.24933760166168212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.3125823974609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.4042111873626709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.5581439971923828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,power_law_1.2,0.7371647834777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,power_law_1.2,0.9499584197998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.2,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.2,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.2,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.2,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.2,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.2,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.2,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.2,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.2,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.2,0.07813760042190551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.2,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.2,0.12718720436096193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.2,0.1517184019088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.2,0.20933759212493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.2,0.21699841022491456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.2,0.2752831935882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.2,0.361023998260498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.2,0.49582719802856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.2,0.6595712184906006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,balanced,0.05028266708056132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,balanced,0.060234665870666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,balanced,0.060122668743133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,balanced,0.0641546646753947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,balanced,0.06474666794141133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,balanced,0.07031466563542683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,balanced,0.08274133503437042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,balanced,0.08876267075538635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,balanced,0.11763200163841248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,balanced,0.13346667091051737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,balanced,0.18000000715255737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,balanced,0.17405333121617636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,balanced,0.22353067000706991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,balanced,0.2469173272450765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,balanced,0.33170666297276813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,balanced,0.39709333578745526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,balanced,0.10069866975148518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,balanced,0.15402133266131082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,balanced,0.18013334274291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,balanced,0.18178667624791464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,balanced,0.18314667542775473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,balanced,0.18402133385340372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,balanced,0.18526933590571085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,balanced,0.1864853302637736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,balanced,0.1901386578877767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,balanced,0.19222933053970337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,balanced,0.19418134291966757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,balanced,0.2018346587816874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,balanced,0.20480533440907797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,balanced,0.21337066094080606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,balanced,0.22566399971644083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,balanced,0.2391093373298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,balanced,0.28731733560562134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,balanced,0.29578665892283124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,balanced,0.41068800290425617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,balanced,0.42607998847961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,balanced,0.6167413393656412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,balanced,0.7039626439412435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,balanced,1.037285327911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,balanced,1.261093298594157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,power_law_1.01,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,power_law_1.01,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,power_law_1.01,0.033932799100875856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,power_law_1.01,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,power_law_1.01,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,power_law_1.01,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,power_law_1.01,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,power_law_1.01,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,power_law_1.01,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,power_law_1.01,0.13103359937667847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,power_law_1.01,0.1759168028831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,power_law_1.01,0.22281599044799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,power_law_1.01,0.28567678928375245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,power_law_1.2,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,power_law_1.2,0.07941759824752807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,power_law_1.2,0.08506879806518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,power_law_1.2,0.09509119987487794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,power_law_1.2,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,power_law_1.2,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,power_law_1.2,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,power_law_1.2,0.09527680277824402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,power_law_1.2,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,power_law_1.2,0.1041599988937378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,power_law_1.2,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,power_law_1.2,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,power_law_1.2,0.12030719518661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,power_law_1.2,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,power_law_1.2,0.13146239519119263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,power_law_1.2,0.15410560369491577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,power_law_1.2,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,power_law_1.2,0.20154240131378173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,power_law_1.2,0.22537600994110107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,power_law_1.2,0.2948607921600342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,power_law_1.2,0.36255359649658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,power_law_1.2,0.48876161575317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,power_law_1.2,0.6326655864715576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,power_law_1.2,0.8569215774536133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,power_law_1.2,1.1890624046325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.2,0.06096000075340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.2,0.06340479850769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.2,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.2,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.2,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.2,0.06709759831428527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.2,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.2,0.08140159845352173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.2,0.10113279819488526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.2,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.2,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.2,0.1711743950843811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.2,0.23300480842590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.2,0.2615423917770386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.2,0.337939190864563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.2,0.45969281196594236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.2,0.61080961227417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.2,0.8748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,balanced,0.04640533526738485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,balanced,0.04613333443800608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,balanced,0.05065066615740458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,balanced,0.06169066826502482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,balanced,0.061290666460990906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,power_law_1.01,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,power_law_1.01,0.10007679462432861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,power_law_1.01,0.12074240446090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,power_law_1.01,0.12202880382537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,power_law_1.01,0.15135999917984008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,power_law_1.01,0.156550395488739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,power_law_1.01,0.15218559503555298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,power_law_1.01,0.1555456042289734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,power_law_1.01,0.15790079832077025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,power_law_1.01,0.1614848017692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,power_law_1.01,0.1568127989768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,power_law_1.01,0.1620352029800415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,power_law_1.01,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,power_law_1.01,0.17696640491485596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,power_law_1.01,0.1769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,power_law_1.01,0.1988800048828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,power_law_1.01,0.2057215929031372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,power_law_1.01,0.23109118938446044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,power_law_1.01,0.2619647979736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,power_law_1.01,0.31987199783325193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,power_law_1.01,0.33393919467926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,power_law_1.01,0.4340735912322998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,power_law_1.01,0.5318399906158447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,power_law_1.01,0.7002431869506835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,power_law_1.01,0.8798912048339844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,balanced,0.07211199899514516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.08637440204620361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.09820799827575684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.11363840103149414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.13040640354156494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.14763519763946534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.1814144015312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.22277119159698486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.3146368026733398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.36745600700378417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,power_law_1.2,0.5604864120483398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,power_law_1.2,0.7013440132141113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,balanced,0.08516266942024231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,balanced,0.08673600355784099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,balanced,0.0972106655438741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,balanced,0.10359999537467957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,balanced,0.12008000413576762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,balanced,0.13674666484196982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,balanced,0.17283199230829874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,balanced,0.2035520076751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,balanced,0.2610879937807719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,balanced,0.32451732953389484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,balanced,0.438480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,balanced,0.5563146670659384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,power_law_1.2,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,power_law_1.2,0.09326080083847046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,power_law_1.2,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,power_law_1.2,0.12359039783477783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,power_law_1.2,0.10485119819641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,power_law_1.2,0.11866240501403809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,power_law_1.2,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,power_law_1.2,0.1283776044845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,power_law_1.2,0.12596479654312134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,power_law_1.2,0.1336127996444702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,power_law_1.2,0.13192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,power_law_1.2,0.1295040011405945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,power_law_1.2,0.13871999979019164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,power_law_1.2,0.14072959423065184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,power_law_1.2,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,power_law_1.2,0.14781440496444703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,power_law_1.2,0.16636799573898314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.2,0.17260160446166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.2,0.19760639667510987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.2,0.22686080932617186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.2,0.2891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.2,0.33056640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.2,0.4937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.2,0.5971392154693603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.2,0.9724287986755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.2,1.2209856033325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,power_law_1.01,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,power_law_1.01,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,power_law_1.01,0.09339519739151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,power_law_1.01,0.10246399641036988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,power_law_1.01,0.11867519617080688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,power_law_1.01,0.13477760553359985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,power_law_1.01,0.1388416051864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,power_law_1.01,0.145798397064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,power_law_1.01,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,power_law_1.01,0.15589120388031005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,power_law_1.01,0.15140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,power_law_1.01,0.16026240587234497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,power_law_1.01,0.1642624020576477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,power_law_1.01,0.17120640277862548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,power_law_1.01,0.18259840011596679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,power_law_1.01,0.2009216070175171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.01,0.21639680862426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.01,0.2540031909942627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.01,0.29295361042022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.01,0.3667263984680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.01,0.4413951873779297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.01,0.5866112232208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.01,0.740505599975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.01,1.0910335540771485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.01,1.3089023590087892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,power_law_1.2,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,power_law_1.2,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,power_law_1.2,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,power_law_1.2,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,power_law_1.2,0.11351679563522339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,power_law_1.2,0.11518720388412476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,power_law_1.2,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,power_law_1.2,0.12399359941482543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,power_law_1.2,0.12815359830856324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,power_law_1.2,0.13104000091552734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,power_law_1.2,0.1363136053085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,power_law_1.2,0.1378368020057678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,power_law_1.2,0.14470399618148805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,power_law_1.2,0.15674240589141847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,power_law_1.2,0.1627519965171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,power_law_1.2,0.16992640495300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,power_law_1.2,0.2006592035293579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,power_law_1.2,0.21763839721679687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,power_law_1.2,0.2711616039276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,power_law_1.2,0.3101056098937988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,power_law_1.2,0.41304960250854494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,power_law_1.2,0.507206392288208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15021439790725707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.2526015996932983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.44525442123413084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.6773375988006591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,power_law_1.2,0.6915264129638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,power_law_1.2,0.912934398651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.653971290588379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.7740800857543946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,power_law_1.2,1.8328447341918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,power_law_1.2,1.8707456588745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,power_law_1.2,1.9449216842651367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.0270208358764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.0572927474975584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.0912384033203124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.111884880065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.16888313293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.2201919555664062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.33372802734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,power_law_1.2,2.465235137939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,power_law_1.2,2.7475135803222654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,power_law_1.2,3.1784767150878905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,power_law_1.2,3.258272171020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,power_law_1.2,4.009920120239258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,power_law_1.2,4.0045631408691404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,power_law_1.2,0.8852671623229981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,power_law_1.2,4.775199890136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,power_law_1.2,6.431104278564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,power_law_1.2,1.2623167991638184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,power_law_1.2,8.021683502197266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,power_law_1.2,8.840211486816406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,power_law_1.2,1.6770559310913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.21106560230255128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.3176512002944946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.40488319396972655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.7241536140441894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.6856319904327393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.7182144165039063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.7386047840118408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.7523647785186768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.7701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.7902527809143066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,power_law_1.01,0.7813055992126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,power_law_1.01,0.8400511741638184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,power_law_1.01,0.8247743606567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,power_law_1.01,0.8343999862670899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,power_law_1.01,0.8826239585876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,power_law_1.01,0.9027647972106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,power_law_1.01,1.0059519767761231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,power_law_1.01,1.1280384063720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,power_law_1.01,1.2732288360595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,power_law_1.01,1.5310272216796874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,power_law_1.01,1.7277631759643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,power_law_1.01,2.2483327865600584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,power_law_1.01,2.8250944137573244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,power_law_1.01,3.9983871459960936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,power_law_1.01,5.13994255065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,power_law_1.01,7.512563323974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,balanced,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,balanced,0.08557867010434468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,balanced,0.08706667025883992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,balanced,0.08756267031033833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,balanced,0.08758399883906047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,balanced,0.08895466725031535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,balanced,0.09103467067082723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,balanced,0.09131733576456706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,balanced,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,balanced,0.10187733173370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,balanced,0.10475200414657593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,balanced,0.10910399754842122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,balanced,0.12219732999801636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,balanced,0.12869333227475485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,balanced,0.14713066816329956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,balanced,0.16665066281954447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,balanced,0.21202667554219565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,balanced,0.24152000745137533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,balanced,0.3378666639328003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,balanced,0.4130773146947225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,balanced,0.5995253324508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,balanced,0.7634080251057943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,power_law_1.01,9.947968292236329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,power_law_1.01,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,power_law_1.01,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,power_law_1.01,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,power_law_1.01,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.01,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.01,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.01,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.01,0.13879679441452025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.01,0.16947840452194213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.01,0.23319680690765382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.01,0.3024447917938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.01,0.43134078979492185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.01,0.535532808303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,balanced,0.1202293336391449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,balanced,0.21183466911315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,balanced,0.20587732394536337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,balanced,0.204202671845754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,balanced,0.20535467068354288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,balanced,0.20546666781107584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,balanced,0.2071626583735148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,balanced,0.20779200394948324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,balanced,0.2073813279469808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,balanced,0.2150933345158895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,balanced,0.21311465899149576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,balanced,0.2140106757481893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,balanced,0.22655999660491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,balanced,0.2509920001029968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,balanced,0.25526400407155353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,balanced,0.24551467100779215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,balanced,0.2995733420054118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,balanced,0.3073119918505351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,balanced,0.3852533499399821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,balanced,0.41764267285664874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,balanced,0.567845344543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,balanced,0.6969866752624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,balanced,0.9958453178405762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,balanced,1.2765173117319744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,balanced,1.8919787406921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,balanced,2.511061350504557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,balanced,0.07605333129564922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,balanced,0.1019040048122406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,balanced,0.15516799688339233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,balanced,0.25280000766118366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,balanced,0.45099735260009766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,balanced,0.6501013437906901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,balanced,0.6549280087153116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,balanced,0.6627093156178793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,balanced,0.6639306545257568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,balanced,0.6695146560668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,balanced,0.6785919666290283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,balanced,0.6756959756215414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,balanced,0.6886080106099447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,balanced,0.697551965713501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,balanced,0.7101813157399496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,balanced,0.7322026888529459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,balanced,0.7740960121154785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,balanced,0.8051839669545492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,balanced,0.8705493609110514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,balanced,0.9454293251037598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,balanced,1.0860319932301838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,balanced,1.276858647664388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,balanced,1.577642599741618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,balanced,1.9930133819580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,balanced,2.5759946505228677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,balanced,3.4572372436523438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.01,0.10067199468612671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.01,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.01,0.09492480158805847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.01,0.09873920083045959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.01,0.09857919812202454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.01,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.01,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.01,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.01,0.10295679569244384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.01,0.1041856050491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.01,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.01,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.01,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.01,0.12339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.01,0.14328960180282593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.01,0.14944000244140626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.01,0.1895616054534912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.01,0.21027839183807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.01,0.27411839962005613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.01,0.26150400638580323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.01,0.36368000507354736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.01,0.4417856216430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.01,0.5888319969177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.01,0.680467176437378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,balanced,0.042165334026018776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,balanced,0.045461331804593406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,balanced,0.04644800225893656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,balanced,0.04746133089065552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,balanced,0.05671466886997223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,balanced,0.06182933350404104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,balanced,0.0642986645301183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,balanced,0.08052266637484233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,balanced,0.09752532839775085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,balanced,0.12214932839075725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,balanced,0.17030932505925497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,balanced,0.20493866999944052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,balanced,0.28147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,balanced,0.3565066655476888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.2,0.03033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.2,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.2,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.2,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.2,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.2,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.2,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.2,0.173580801486969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.2,0.22417919635772704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.2,0.31811840534210206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,power_law_1.01,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,power_law_1.01,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,power_law_1.01,0.09426559805870056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,power_law_1.01,0.10158079862594604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,power_law_1.01,0.11944320201873779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,power_law_1.01,0.13911039829254152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,power_law_1.01,0.17219200134277343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,power_law_1.01,0.21511681079864503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,power_law_1.01,0.2778496026992798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,power_law_1.01,0.3546751976013184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,power_law_1.01,0.4832320213317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,power_law_1.01,0.6581759929656983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,balanced,0.06804266571998596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,balanced,0.0555626650651296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,balanced,0.05538133283456167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,balanced,0.054805333415667214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,balanced,0.05450133482615153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,balanced,0.055120001236597695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,balanced,0.05783466498057047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,balanced,0.06705600023269653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,balanced,0.09711466232935588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,balanced,0.10312533378601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,balanced,0.12181333700815837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,balanced,0.14941333731015524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,balanced,0.19086400667826334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,balanced,0.22613332668940225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,balanced,0.31492799520492554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,balanced,0.3901066780090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.030720001459121703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.09242240190505982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,balanced,0.0799786647160848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,balanced,0.09112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,balanced,0.09454400340716045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,balanced,0.0936906635761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,balanced,0.09325866897900899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,balanced,0.09412266810735066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,balanced,0.09691733121871948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,balanced,0.09711466232935588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,balanced,0.099589337905248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,balanced,0.09914132952690125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,balanced,0.10322133700052898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,balanced,0.10572800040245056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,balanced,0.11967466274897258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,balanced,0.13036800424257913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,balanced,0.1574986676375071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,balanced,0.17228267590204874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,balanced,0.2206773360570272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,balanced,0.22105065981547037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,balanced,0.2931999961535136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,balanced,0.3259893258412679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,balanced,0.45501867930094403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,balanced,0.5373280048370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.12498559951782226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.1553536057472229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.20332798957824708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.2518079996109009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.07308800220489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.12088960409164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.15576959848403932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.1551103949546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.20354559421539306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.24828801155090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.33581440448760985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.40855679512023924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.5724095821380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.42675199508666994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.8045951843261718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,power_law_1.2,0.37957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,balanced,0.03815466662247976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,balanced,0.04289066791534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,balanced,0.04612799982229868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,balanced,0.06990399956703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,balanced,0.10135466853777568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,balanced,0.12994666894276938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,balanced,0.1525920033454895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,balanced,0.21606934070587158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,balanced,0.2635519901911418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.06270719766616821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.1126207947731018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.10156160593032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.11013120412826538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.12874879837036132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.14826879501342774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.18099199533462523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.19689600467681884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.2546112060546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.28482561111450194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,power_law_1.01,0.40233597755432127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,power_law_1.01,0.44109439849853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,power_law_1.01,0.6047039985656738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,power_law_1.01,1.0020352363586427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,power_law_1.01,1.3321087837219239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,power_law_1.01,1.572326374053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,balanced,0.07081066568692525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,balanced,0.07835733393828075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,balanced,0.09036266803741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,balanced,0.1181706686814626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,balanced,0.17110933860143027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,balanced,0.22231467564900717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,balanced,0.2130933403968811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,balanced,0.21403199434280396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,balanced,0.21707199017206827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,balanced,0.21738133827845255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,balanced,0.21820799509684244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,balanced,0.22275733947753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,balanced,0.22697067260742188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,balanced,0.23414933681488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,balanced,0.23875200748443604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,balanced,0.2485599915186564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,balanced,0.2723466753959656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.2918933431307475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.3392053445180257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.3768693208694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.46859200795491535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.5623413324356079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.7480053106943766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,balanced,1.0174187024434407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,balanced,1.403322696685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,balanced,1.8618720372517903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.07296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.10039680004119873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.12028800249099732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,power_law_1.01,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.14837759733200073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,power_law_1.01,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.15816320180892945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.17274880409240723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,power_law_1.01,0.14905600547790526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.17760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,power_law_1.01,0.18124159574508666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.18098560571670533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,power_law_1.01,0.26501760482788084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.18201600313186644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,power_law_1.01,0.33238399028778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.19095040559768678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.20277121067047119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,power_law_1.01,0.3879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.20346879959106445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,power_law_1.01,0.4853248119354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.21417601108551027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.22088320255279542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,power_law_1.01,0.4891007900238037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.23256959915161132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,power_law_1.01,0.5040832042694092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.25338239669799806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,power_law_1.01,0.5036352157592774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.282476806640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.33331201076507566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,power_law_1.01,0.5476607799530029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.3855743885040283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,power_law_1.01,0.5473023891448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.49628801345825196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,power_law_1.01,0.5414015769958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.607366418838501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,power_law_1.01,0.5792191982269287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.8378432273864747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,power_law_1.01,0.611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,1.0691136360168456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,power_law_1.01,0.6307263851165772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.5098112106323243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,1.9351999282836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,power_law_1.01,0.6537087917327881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,power_law_1.01,0.6828864097595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,power_law_1.01,0.7929664134979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,power_law_1.01,0.8534272193908692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,power_law_1.01,1.0367936134338378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,power_law_1.01,1.1514752388000489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,power_law_1.01,1.4720447540283204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,power_law_1.01,1.765228843688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,power_law_1.01,2.2851327896118163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,power_law_1.01,2.8038591384887694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,balanced,0.04614399870236715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,balanced,0.0658186674118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,balanced,0.09777067104975383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,balanced,0.14890666802724203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,balanced,0.15219733119010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,balanced,0.15520532925923666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,balanced,0.1541813313961029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,balanced,0.1565013329188029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,balanced,0.15988799929618835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,balanced,0.15915200114250183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,balanced,0.16154666741689047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,balanced,0.162800004084905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,balanced,0.16529599825541177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,balanced,0.17403733730316162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,balanced,0.17645865678787231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,balanced,0.1842986742655436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,balanced,0.20252267519632974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.22241600354512533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.25148266553878784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.28351465861002606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.38020801544189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.425493319829305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.6047946612040201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.7232800324757894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,balanced,1.0296586354573567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.345589319864909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.05815039873123169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.08501120209693909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.11388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.17898240089416503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.2213184118270874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.32362239360809325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.2,0.10327680110931396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.2,0.10658559799194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.2,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.2,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.2,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.2,0.05486080050468445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.2,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.2,0.06995199918746949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.2,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.2,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.2,0.07902079820632935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.3919167995452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.2,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.2,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.2,0.11812479496002197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.2,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.2,0.14684159755706788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.2,0.15676800012588502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.2,0.21091198921203613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.2,0.21687679290771483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.2,0.30467839241027833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.2,0.45343360900878904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.2,0.6788288116455078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,power_law_1.2,0.5737215995788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,power_law_1.2,0.7335936069488526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.2,0.7366720199584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,power_law_1.01,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,power_law_1.01,0.07301120162010193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,power_law_1.01,0.0905344009399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,power_law_1.01,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,power_law_1.01,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,power_law_1.01,0.10373120307922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,power_law_1.01,0.10673279762268066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,power_law_1.01,0.10661760568618775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,power_law_1.01,0.1160256028175354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,power_law_1.01,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,power_law_1.01,0.12698240280151368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,power_law_1.01,0.14689279794692994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,power_law_1.01,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,power_law_1.01,0.19005440473556517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,power_law_1.01,0.22110719680786134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,power_law_1.01,0.28920319080352785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,power_law_1.01,0.32899839878082277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,power_law_1.01,0.4651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,power_law_1.01,0.5628416061401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,power_law_1.01,0.8171839714050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,power_law_1.01,1.0822527885437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,balanced,0.04800533254941305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,balanced,0.05045333504676819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,balanced,0.06163200239340464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,balanced,0.07375466823577881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,balanced,0.07840533554553986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,balanced,0.08906666437784831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,balanced,0.09699199597040813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,balanced,0.12157332897186279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,balanced,0.14198399583498636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,balanced,0.18963199853897095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,balanced,0.22976533571879068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,balanced,0.32197866837183636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,balanced,0.40305598576863605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,balanced,0.11086933811505635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,balanced,0.12841066718101501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,balanced,0.12914133071899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,balanced,0.12961066762606302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,balanced,0.12928533554077148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,balanced,0.13129066427548727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,balanced,0.13200533390045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,balanced,0.13336533308029175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,balanced,0.13824533422787985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,balanced,0.14364799857139587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,balanced,0.14780799547831217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,balanced,0.15355733036994934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,balanced,0.16219199697176614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,balanced,0.18179200092951456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,balanced,0.19569599628448486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,balanced,0.25204267104466754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,balanced,0.27561599016189575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,balanced,0.38684264818827313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,balanced,0.4344746669133504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,balanced,0.6352266470591227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,balanced,0.7359626293182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,balanced,1.1721866925557454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,balanced,1.373253345489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,balanced,0.029472000896930695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,balanced,0.05927466849486033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,balanced,0.05739733576774597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,balanced,0.05916800101598104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,balanced,0.05860800047715505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,balanced,0.06469333171844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,balanced,0.06543466448783875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,balanced,0.08243733147780101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,balanced,0.08517332871754964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,balanced,0.10821333527565002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,balanced,0.1202133297920227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,balanced,0.16146666804949442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,balanced,0.1684373418490092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,balanced,0.24153600136439005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,balanced,0.2815413276354472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,power_law_1.01,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,power_law_1.01,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,power_law_1.01,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,power_law_1.01,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,power_law_1.01,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,power_law_1.01,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,power_law_1.01,0.0800383985042572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,power_law_1.01,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,power_law_1.01,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,power_law_1.01,0.11606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.01,0.14673279523849486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.01,0.20158720016479492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.01,0.2689215898513794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.01,0.3677759885787964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.01,0.5036799907684326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.01,0.7483136177062988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.01,0.9749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.01,1.4482815742492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.01,1.879737663269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,4,power_law_1.01,0.13889280557632447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,4,power_law_1.01,0.19124480485916137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,4,power_law_1.01,0.17557120323181152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,4,power_law_1.01,0.2531712055206299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,4,power_law_1.01,0.23656320571899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,4,power_law_1.01,0.31836159229278566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,4,power_law_1.01,0.3369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,4,power_law_1.01,0.33496320247650146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,4,power_law_1.01,0.3502336025238037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,4,power_law_1.01,0.353657603263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,4,power_law_1.01,0.36649599075317385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,4,power_law_1.01,0.36618239879608155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,4,power_law_1.01,0.3685695886611938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,4,power_law_1.01,0.3784192085266113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,4,power_law_1.01,0.3762495994567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,4,power_law_1.01,0.39018878936767576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,4,power_law_1.01,0.40592641830444337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,4,power_law_1.01,0.42237439155578616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.01,0.43799681663513185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.01,0.511571216583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.01,0.5628992080688476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.01,0.7620416164398194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.01,0.8422528266906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.01,1.173523235321045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.01,1.4887616157531738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.01,2.371968078613281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.01,2.7018112182617187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.13121919631958007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.09377920031547546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.09404159784317016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.10019840002059936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.10430079698562622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.10572799444198608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.1473536014556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.15022720098495485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.17824000120162964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.20807039737701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.27869439125061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.28818559646606445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,power_law_1.2,0.36871678829193116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,power_law_1.2,0.49735679626464846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,power_law_1.2,0.7909632205963135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,power_law_1.2,0.9806271553039551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,balanced,0.06461333235104878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,balanced,0.06884266436100006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,balanced,0.07089066505432129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,balanced,0.07223466535409291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,balanced,0.07653333246707916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,balanced,0.07427733143170674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,balanced,0.07420266668001811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,balanced,0.07691200077533722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,balanced,0.08255999783674876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,balanced,0.09241599837938945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,balanced,0.09694400429725647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,balanced,0.10810133814811707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,balanced,0.11841600139935811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,balanced,0.1394773324330648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,balanced,0.15785066286722818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,balanced,0.19602133830388388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,balanced,0.23444799582163492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,balanced,0.31353600819905597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,balanced,0.387994647026062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,power_law_1.01,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,power_law_1.01,0.15061119794845582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,power_law_1.01,0.2102720022201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,power_law_1.01,0.27589759826660154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,power_law_1.01,0.3795583963394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,power_law_1.01,0.49718399047851564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,power_law_1.01,0.6577663898468018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,power_law_1.01,0.6466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,power_law_1.01,0.6777088165283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,power_law_1.01,0.7076479911804199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,power_law_1.01,0.7449984073638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,power_law_1.01,0.7400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,power_law_1.01,0.768620777130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,power_law_1.01,0.748518419265747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,power_law_1.01,0.7686912059783936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,power_law_1.01,0.787007999420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,power_law_1.01,0.8401087760925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,power_law_1.01,0.9116288185119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,power_law_1.01,0.9207167625427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,power_law_1.01,1.0350848197937013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,power_law_1.01,1.0741632461547852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,power_law_1.01,1.2575615882873534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,power_law_1.01,1.3984319686889648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,power_law_1.01,1.7818944931030274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,power_law_1.01,2.067923164367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,power_law_1.01,2.7546432495117186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,power_law_1.01,3.42242546081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,balanced,0.07031466563542683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,balanced,0.07203199962774913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,balanced,0.07145066559314728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,balanced,0.0724480003118515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,balanced,0.07223466535409291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,balanced,0.07452799876530965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,balanced,0.07430399954319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,balanced,0.07975466549396515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,balanced,0.08705066641171773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,balanced,0.08932800094286601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,balanced,0.09511466821034749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,balanced,0.10653866330782573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.11577600240707397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.137855996688207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.1602826714515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.21010132630666098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.2641119956970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.35099200407663983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.4213333527247111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.6040159861246744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,balanced,0.7842293580373129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,power_law_1.2,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,power_law_1.2,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,power_law_1.2,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,power_law_1.2,0.06513919830322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,power_law_1.2,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,power_law_1.2,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,power_law_1.2,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,power_law_1.2,0.08535040020942689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,power_law_1.2,0.10362880229949951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.2,0.11584639549255371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.2,0.1408959984779358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.2,0.16335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.2,0.21877119541168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.2,0.25793919563293455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.2,0.37353599071502686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.2,0.4942016124725342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.2,0.9000191688537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.2,1.030355167388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.2,0.025644800066947936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.2,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.2,0.028019198775291444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.2,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.2,0.0298880010843277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.2,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.2,0.028774398565292358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.2,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.2,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.2,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.2,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.2,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.2,0.15319039821624755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,balanced,0.042730664213498436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,balanced,0.07622933387756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,balanced,0.10371733705202739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,balanced,0.10568533341089885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,balanced,0.10756799578666687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,balanced,0.10869866609573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,balanced,0.10963732997576396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,balanced,0.1104853351910909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,balanced,0.11089600125948589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,balanced,0.10942932963371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,balanced,0.11372266213099162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,balanced,0.12136000394821167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,balanced,0.1225440005461375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,balanced,0.12972266475359598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,balanced,0.14260266224543253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.2,0.22506239414215087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,balanced,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,balanced,0.17610132694244385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,balanced,0.19903467098871866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,balanced,0.24555200338363647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,balanced,0.2922079960505168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,balanced,0.39477332433064777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.2,0.28073599338531496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,balanced,0.4814613262812297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,balanced,0.681546688079834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,balanced,0.8634666601816813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.09666560292243957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.11880960464477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.14287999868392945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.1859455943107605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.2400831937789917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.3181119918823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.4161856174468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.09122560024261475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.08784000277519226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.09164159893989562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.11324800252914428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.12294399738311768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.18746880292892457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.203385591506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.25501439571380613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.2895551919937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.3803584098815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.5345344066619873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.7835008144378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,0.9903743743896485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10279040336608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.1319808006286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.15943679809570313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.22439041137695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.24033279418945314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.24528000354766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.2515455961227417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.2656768083572388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.27470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.28170878887176515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.2918207883834839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.30891520977020265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.3183039903640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.3348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.3962111949920654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.4254720211029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5137407779693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.5832767963409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.7396671772003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.9139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.2157888412475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.504800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.095583915710449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.7225727081298827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,balanced,0.0553653339544932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,balanced,0.07728533446788788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,balanced,0.11969600121180217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,balanced,0.1206719974676768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,balanced,0.12077866991360982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,balanced,0.1216266651948293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,balanced,0.12324266632397969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,balanced,0.1239306628704071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,balanced,0.1253546675046285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,balanced,0.12470933794975281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,balanced,0.12745066483815512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,balanced,0.1295146644115448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,balanced,0.1304586629072825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,balanced,0.13671466708183289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,balanced,0.1567093332608541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,balanced,0.156549334526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,balanced,0.15837867061297098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,balanced,0.18613332509994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,balanced,0.19827200969060263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,balanced,0.24700266122817993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,balanced,0.28117867310841876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,balanced,0.3837386767069499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,balanced,0.4777173201243083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,balanced,0.6747840245564779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,balanced,0.8688266277313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,balanced,1.2703839937845867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,balanced,1.669541358947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.5300543785095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,balanced,0.09760533769925435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,balanced,0.14988266428311667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,balanced,0.2544800043106079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,balanced,0.46663467089335126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,balanced,0.8842026392618815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,balanced,1.0941173235575359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,balanced,1.1005760033925374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,balanced,1.1053866545359294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,balanced,1.1093333562215169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,balanced,1.1141706307729085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,balanced,1.1213866869608562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,balanced,1.1273919741312664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,balanced,1.1342453161875408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,balanced,1.1471306482950847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,balanced,1.160965363184611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,balanced,1.170026699701945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,balanced,1.1951200167338054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,balanced,1.2784639994303386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,balanced,1.272005319595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,balanced,1.5827253659566243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,balanced,1.4472479820251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,balanced,2.2036959330240884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,balanced,1.8033547401428223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,balanced,2.7653331756591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,balanced,2.8221972783406577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,balanced,4.54747200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,balanced,5.124874750773112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,balanced,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,balanced,0.04470400015513102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,balanced,0.04827733337879181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,balanced,0.05541333556175232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,balanced,0.07019199927647908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,balanced,0.07659733295440674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,balanced,0.10520000259081523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,balanced,0.11807466546694438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,balanced,0.1511520047982534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,balanced,0.176639993985494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,balanced,0.2249386707941691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,balanced,0.2726186712582906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,balanced,0.3674986759821574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,balanced,0.46133867899576825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,power_law_1.01,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,power_law_1.01,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,power_law_1.01,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,power_law_1.01,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,power_law_1.01,0.08008319735527039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,power_law_1.01,0.10474879741668701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,power_law_1.01,0.131660795211792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,power_law_1.01,0.15326720476150513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,power_law_1.01,0.19723520278930665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,power_law_1.01,0.23728640079498292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,power_law_1.01,0.3331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,power_law_1.01,0.41475839614868165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,power_law_1.01,0.6210879802703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,power_law_1.01,0.7792448043823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,balanced,0.03496533383925756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,balanced,0.03624533365170161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,balanced,0.0391839991013209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,balanced,0.052383999029795326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,balanced,0.0626453310251236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,balanced,0.09407466650009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,balanced,0.11558399597803752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,balanced,0.13381333152453104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,balanced,0.19620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,balanced,0.2334666649500529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,balanced,0.32892799377441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,balanced,0.4300160010655721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,power_law_1.01,0.09993600249290466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,power_law_1.01,0.15112320184707642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,power_law_1.01,0.2548608064651489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,power_law_1.01,0.401043176651001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,power_law_1.01,0.5284351825714111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,power_law_1.01,0.8514240264892579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,power_law_1.01,0.9186495780944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,power_law_1.01,0.9575296401977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,power_law_1.01,1.0001536369323731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,power_law_1.01,0.9752256393432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,power_law_1.01,1.0113920211791991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,power_law_1.01,1.0374719619750976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,power_law_1.01,1.037548828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,power_law_1.01,1.1044992446899413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,power_law_1.01,1.1380736351013183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,power_law_1.01,1.1579008102416992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,power_law_1.01,1.1840319633483887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,power_law_1.01,1.2788607597351074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,power_law_1.01,1.3445504188537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,power_law_1.01,1.569542407989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,power_law_1.01,1.7316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,power_law_1.01,2.077497673034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,power_law_1.01,2.1951616287231444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,power_law_1.01,2.6903360366821287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,power_law_1.01,3.5921985626220705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,power_law_1.01,4.592038345336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,power_law_1.01,5.560281753540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.1987391948699951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.20172159671783446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.20280959606170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.20177919864654542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.20935680866241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.2133631944656372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.21413118839263917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.21578879356384278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.22646400928497315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.228985595703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.23609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.24423038959503174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.26623361110687255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.26987519264221194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.3206720113754272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.34226560592651367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,power_law_1.01,0.4077631950378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,power_law_1.01,0.483673620223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,power_law_1.01,0.6373504161834717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,power_law_1.01,0.7910079956054688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,power_law_1.01,1.0782912254333497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,power_law_1.01,1.401260757446289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,power_law_1.01,2.070931243896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,power_law_1.01,2.6676992416381835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.1626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.20860159397125244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.26343679428100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.3355391979217529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.3091007947921753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.49689598083496095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.5157760143280029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.5068480014801026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.5172416210174561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.5314752101898194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.5484288215637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.545804786682129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.5851200103759766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.5495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.5926591873168945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.5907711982727051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.6609216213226319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,balanced,0.05229333539803823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,power_law_1.2,0.7556287765502929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,balanced,0.11800000071525574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,balanced,0.11904000242551167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,balanced,0.12270399928092957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,balanced,0.1223413348197937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,power_law_1.2,0.7693056106567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,balanced,0.12166933218638103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,balanced,0.12242133418718974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,balanced,0.12332800030708313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,balanced,0.12402133146921794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,balanced,0.12589333454767862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,power_law_1.2,0.9426495552062988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,balanced,0.13353600104649863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,balanced,0.1357493301232656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,balanced,0.1386133333047231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,balanced,0.1514079968134562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.16038399934768677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,power_law_1.2,0.8966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.17923200130462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.20586667458216348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.24254933993021646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.2722933292388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.3635626633961995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.42202667395273846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.597765326499939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.7427679697672526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,power_law_1.2,1.0430527687072755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,power_law_1.2,1.2535360336303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,power_law_1.2,1.4769536018371583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,power_law_1.2,1.9062143325805665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,power_law_1.2,2.4093631744384765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,power_law_1.2,3.227372741699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,balanced,0.07504533231258392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,balanced,0.11400533715883891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,balanced,0.17841599384943643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,balanced,0.30854399998982746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,balanced,0.30851733684539795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,balanced,0.3116053342819214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,balanced,0.3130506674448649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,balanced,0.3118506669998169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,balanced,0.31461866696675617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,balanced,0.322869340578715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,balanced,0.3213866750399272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,balanced,0.3264639973640442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,balanced,0.3336533308029175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,balanced,0.3375946680704753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,balanced,0.3463199933369954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,balanced,0.3700266679128011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,balanced,0.3856426477432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,balanced,0.43163732687632245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,balanced,0.4768213431040446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,balanced,0.5868373314539591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,balanced,0.6463840007781982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,balanced,0.8782453536987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,balanced,1.0331786473592122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,balanced,1.4603840510050456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,balanced,1.829312006632487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.10792319774627686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.1428096055984497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.1893056035041809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.1865023970603943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.27375359535217286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.2827071905136108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.27793920040130615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.28254721164703367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.285478401184082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.29687039852142333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.2919552087783813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.2969088077545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.3021888017654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.3243839979171753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.34112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.3330944061279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.3771712064743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.40453758239746096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.4840703964233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.5200128078460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,power_law_1.01,0.635532808303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,power_law_1.01,0.6766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,power_law_1.01,0.7768959999084473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,power_law_1.01,0.9731648445129395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,power_law_1.01,1.3562047958374024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,power_law_1.01,1.7316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.10365439653396606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.12140159606933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.15676800012588502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.07147520184516906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.0835968017578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.09111040234565734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.10926719903945922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.16431360244750975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.2239936113357544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.27139840126037595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.3353919982910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,power_law_1.2,0.44739837646484376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.20862081050872802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.23837440013885497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,power_law_1.2,0.6788352012634278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.36248319149017333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.01,0.19440640211105348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.01,0.32592000961303713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.01,0.4038976192474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.01,0.6410687923431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.01,0.8098560333251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.01,1.271763229370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.01,1.2878656387329102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.01,1.3285440444946288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.01,1.4005632400512695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.01,1.4134336471557618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.01,1.4116928100585937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.01,1.4565759658813477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.01,1.508563232421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.01,1.5224127769470215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.01,1.5604415893554688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.01,1.584812831878662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.01,1.679142379760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.01,1.8439359664916992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.01,1.9069248199462892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.01,2.1150463104248045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.01,2.259891128540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.01,2.7301055908203127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.01,2.8952640533447265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.01,3.5899326324462892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.01,4.428096008300781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.01,5.611321640014649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.01,6.398988723754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.0876800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.13880319595336915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.18142720460891723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.074508798122406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.08744320273399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.09032959938049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.08194559812545776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.09070720076560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.09823359847068787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.09699199795722961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.11379200220108032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.14608639478683472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.20517120361328126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.24825599193572997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.31138560771942136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.4032383918762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,power_law_1.01,0.5324863910675048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,power_law_1.01,0.6608960151672363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,balanced,0.06182399888833364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,balanced,0.06323199967543285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,balanced,0.06162666777769724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,balanced,0.06538133323192596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,balanced,0.06460266808668773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,balanced,0.06446399788061778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,balanced,0.07081600030263265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,balanced,0.09327999750773112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,balanced,0.09594666957855225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,balanced,0.11956266562143962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,balanced,0.1325813333193461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,balanced,0.17697066068649292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,balanced,0.18571732441584268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,balanced,0.25091199080149335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,balanced,0.3062079946200053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,balanced,0.1341653366883596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,balanced,0.12113599975903828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,balanced,0.21727999051411948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,balanced,0.2154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,balanced,0.21626667181650797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,balanced,0.21792533000310263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,balanced,0.2183199922243754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,balanced,0.21565866470336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,balanced,0.2163626750310262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,balanced,0.21754133701324463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,balanced,0.225765327612559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,balanced,0.22486400604248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,balanced,0.22489066918691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,balanced,0.2363306681315104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,balanced,0.25650666157404584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,balanced,0.2570880055427551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,balanced,0.2487199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,balanced,0.3173866669336955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,balanced,0.2927146752675374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,balanced,0.41369601090749103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,balanced,0.41105600198109943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,balanced,0.5490506490071615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,balanced,0.68067733446757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,balanced,0.9504799842834473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,balanced,1.2679466406504314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,balanced,1.821440060933431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,balanced,2.4240479469299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,balanced,0.033941333492596946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,balanced,0.03483733286460241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,balanced,0.038736000657081604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,balanced,0.05969599882761637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,balanced,0.09685333569844563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,balanced,0.11480533083279927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,balanced,0.16033599774042764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,balanced,0.19448532660802206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,balanced,0.26682666937510174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,balanced,0.3442293405532837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.048127999901771544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.06997119784355163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.11260800361633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.12934399843215943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.1470080018043518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.1829568028450012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.22492799758911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.2777087926864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.37988479137420655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.4760704040527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,power_law_1.2,0.6443136215209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,power_law_1.2,0.8285120010375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,power_law_1.2,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,power_law_1.2,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,power_law_1.2,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,power_law_1.2,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,power_law_1.2,0.07015039920806884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,power_law_1.2,0.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,power_law_1.2,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,power_law_1.2,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,power_law_1.2,0.08968960046768189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,power_law_1.2,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,power_law_1.2,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,power_law_1.2,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,power_law_1.2,0.12772480249404908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,power_law_1.2,0.17543679475784302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,power_law_1.2,0.2244352102279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,power_law_1.2,0.3122688055038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,power_law_1.2,0.3448767900466919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,power_law_1.2,0.4891392230987549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,power_law_1.2,0.6171840190887451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,power_law_1.2,0.9704832077026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,power_law_1.2,1.3296768188476562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,power_law_1.01,0.022540800273418427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,power_law_1.01,0.023340800404548646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,power_law_1.01,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,power_law_1.01,0.02359039932489395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,power_law_1.01,0.022489599883556366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,power_law_1.01,0.02465279996395111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,power_law_1.01,0.02507520020008087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,power_law_1.01,0.02624639868736267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,power_law_1.01,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,power_law_1.01,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,power_law_1.01,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,power_law_1.01,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,power_law_1.01,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.09103999733924865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.11192320585250855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.13717759847640992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,power_law_1.2,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,power_law_1.2,0.02661119997501373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,power_law_1.2,0.026707199215888978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,power_law_1.2,0.029356798529624938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,power_law_1.2,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,power_law_1.2,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,power_law_1.01,0.026681599020957947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,power_law_1.01,0.02590720057487488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,power_law_1.01,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,power_law_1.01,0.02810879945755005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,power_law_1.01,0.028992000222206115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,power_law_1.01,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,power_law_1.01,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,power_law_1.01,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,power_law_1.01,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,power_law_1.01,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,power_law_1.01,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,power_law_1.01,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,power_law_1.01,0.07402240037918091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,power_law_1.01,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,power_law_1.01,0.11484160423278808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,power_law_1.01,0.13696000576019288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,power_law_1.01,0.21118719577789308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,power_law_1.01,0.2951296091079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,power_law_1.01,0.39188480377197266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,power_law_1.01,0.48061442375183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,power_law_1.2,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,power_law_1.2,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,power_law_1.2,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,power_law_1.2,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,power_law_1.2,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,power_law_1.2,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,power_law_1.2,0.11204479932785034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,power_law_1.2,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,power_law_1.2,0.21510400772094726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,power_law_1.2,0.29395840167999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,power_law_1.2,0.3780607938766479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,power_law_1.2,0.6182144165039063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,power_law_1.2,0.8760767936706543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.02887679934501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.028972798585891725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.030649599432945252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.0626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.13217920064926147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.1460543990135193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.21289598941802979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,power_law_1.2,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,power_law_1.2,0.08020480275154114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,power_law_1.2,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,power_law_1.2,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,power_law_1.2,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,power_law_1.2,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,power_law_1.2,0.08630399703979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,power_law_1.2,0.08811519742012024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.2644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,power_law_1.2,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,power_law_1.2,0.08881279826164246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,power_law_1.2,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,power_law_1.2,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,power_law_1.2,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,power_law_1.2,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.2,0.1297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.2,0.16524159908294678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.2,0.19447679519653321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.2,0.23805439472198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.2,0.30628480911254885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.2,0.4702911853790283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.2,0.6116928100585938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.2,0.9210495948791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.2,1.2942527770996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,power_law_1.2,0.42691841125488283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.08251519799232483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.12120959758758545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.15022079944610595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.2286207914352417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.2703999996185303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,power_law_1.01,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,power_law_1.01,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,power_law_1.01,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,power_law_1.01,0.04530560076236725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,power_law_1.01,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,power_law_1.01,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,power_law_1.01,0.07594879865646362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,power_law_1.01,0.09079679846763611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.01,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.01,0.14554879665374756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.01,0.15970560312271118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.01,0.2342655897140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.01,0.35330560207366946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.01,0.47166080474853517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,balanced,0.03213333338499069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,balanced,0.03312533348798752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,balanced,0.035018667578697205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,balanced,0.07966933151086171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,balanced,0.0798880010843277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,balanced,0.07937600215276082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,balanced,0.08021333316961925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,balanced,0.08252266546090443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,balanced,0.08221866687138875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,balanced,0.08550399541854858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,balanced,0.08507200082143147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,balanced,0.09103467067082723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,balanced,0.09186133742332458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,balanced,0.09539733330408733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,balanced,0.10063466429710388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,balanced,0.11421333750089009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,balanced,0.13150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,balanced,0.135343998670578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,balanced,0.1677066683769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,balanced,0.18651199340820312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,balanced,0.25996265808741253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,balanced,0.3054826656977336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,balanced,0.4176106850306193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,balanced,0.5282933314641317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.01,0.5858240127563477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.01,0.9446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,power_law_1.2,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,power_law_1.2,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,power_law_1.2,0.07811200022697448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,power_law_1.2,0.07970560193061829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,power_law_1.2,0.08202239871025085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,power_law_1.2,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,power_law_1.2,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,power_law_1.2,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,power_law_1.2,0.0953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,power_law_1.2,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,power_law_1.2,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,power_law_1.2,0.11068799495697021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,power_law_1.2,0.13322880268096923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,power_law_1.2,0.14914560317993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,power_law_1.2,0.1873088002204895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,power_law_1.2,0.22561280727386473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,power_law_1.2,0.30283520221710203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,power_law_1.2,0.3613951921463013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,power_law_1.2,0.5055103778839112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,power_law_1.2,0.6426496028900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.01,1.179315185546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,power_law_1.2,0.9206080436706543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,balanced,0.050442665815353394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,balanced,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,balanced,0.0710346649090449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,balanced,0.07234666744867961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,balanced,0.07356266677379608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,balanced,0.07361599802970886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,balanced,0.07864533364772797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,balanced,0.07856533428033192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,balanced,0.07720533510049184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,balanced,0.08450133601824443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,balanced,0.08900800347328186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,balanced,0.10410666465759277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,balanced,0.11339199542999268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,balanced,0.14257599910100302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,balanced,0.16110400358835855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,balanced,0.21408534049987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,balanced,0.20068800449371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,balanced,0.2584480047225952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,balanced,0.30425065755844116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,balanced,0.41522665818532306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,balanced,0.49538131554921466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,power_law_1.2,1.1810303688049317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,balanced,0.05239999790986379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,balanced,0.07069866855939229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,balanced,0.08963200449943542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,balanced,0.09341333309809367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,balanced,0.09114133318265279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,balanced,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,balanced,0.0932373305161794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,balanced,0.09310400485992432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,balanced,0.09684800108273824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,balanced,0.10693333546320598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,balanced,0.10946666200955708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,balanced,0.11146666606267293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,balanced,0.12358400225639343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,balanced,0.1302720010280609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,balanced,0.14622933665911356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,balanced,0.16280532876650491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,balanced,0.20572799444198608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,balanced,0.24457067251205444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,balanced,0.35372265179951984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,balanced,0.4412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,balanced,0.641701340675354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,balanced,0.8259413242340088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,balanced,0.05965333183606466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,balanced,0.06576533118883769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,balanced,0.06206400195757548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,balanced,0.06257600088914235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,balanced,0.06582933167616527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,balanced,0.0680320014556249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,balanced,0.07106133302052815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,balanced,0.08082133531570435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,balanced,0.08679466446240743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,balanced,0.1102186640103658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,balanced,0.13360533118247986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,balanced,0.16781334082285562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,balanced,0.20087466637293497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,balanced,0.2675253351529439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,balanced,0.3320746620496114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,balanced,0.4642080068588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,balanced,0.5910666783650717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.01,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.01,0.09671040177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.01,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.01,0.103603196144104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.01,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.01,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.01,0.10145920515060425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.01,0.10427520275115967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.01,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.01,0.10792959928512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.01,0.11161600351333618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.01,0.12026879787445069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.01,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.01,0.1268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.01,0.158297598361969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.01,0.19185279607772826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.01,0.23517439365386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.01,0.29560959339141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.01,0.3128704071044922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.01,0.40931200981140137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.01,0.49776639938354494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.01,0.7275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.01,0.8583423614501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,balanced,0.041189332803090416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,balanced,0.04312000175317129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,balanced,0.055200000603993736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,balanced,0.07390399773915608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.08469866712888081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.1013866662979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.12096533179283142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.16241066654523215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.19092265764872232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.27164266506830853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.34434131781260174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.494213342666626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.6390399932861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.1563647985458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,balanced,0.10950400431950887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,balanced,0.14454399545987448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,balanced,0.2079626719156901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,balanced,0.3181653420130412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,balanced,0.4306346575419108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,balanced,0.4312853415807088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,balanced,0.4320746660232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,balanced,0.433135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,balanced,0.43516798814137775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,balanced,0.4469386736551921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,balanced,0.45048534870147705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,balanced,0.4510186513264974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,balanced,0.4617706537246704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,balanced,0.46486934026082355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,balanced,0.4840533336003621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,balanced,0.5272746483484904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.5609386761983236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,balanced,0.6377653280893961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,balanced,0.7069546381632487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,balanced,0.8673120339711508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,balanced,1.0196426709493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.20881919860839843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,balanced,1.3300533294677734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,balanced,1.8059679667154949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,balanced,2.451082706451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,balanced,3.271615982055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.2441728115081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,power_law_1.01,0.35798399448394774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,power_law_1.01,0.44958081245422366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.14099839925765992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.1916607975959778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.2518784046173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.3456063985824585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,balanced,0.04833599925041199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,balanced,0.053301334381103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,balanced,0.07701866825421651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,balanced,0.10247466961542766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,balanced,0.13634133338928223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,balanced,0.15335466464360556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,balanced,0.15587733189264932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,balanced,0.15599466363588968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,balanced,0.15558399756749472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,balanced,0.1558133363723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,balanced,0.15743999679883322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,balanced,0.1579146683216095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,balanced,0.16034133235613504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,balanced,0.1609386702378591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,balanced,0.16487466295560202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,balanced,0.16782933473587036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,balanced,0.17137600978215536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,balanced,0.19681066274642944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,balanced,0.19126399358113608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,balanced,0.24275734027226767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,balanced,0.23340266942977905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,balanced,0.3495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,balanced,0.29785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,balanced,0.4152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,balanced,0.42154133319854736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,balanced,0.5978613297144572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,balanced,0.6642453273137411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.1355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.1783360004425049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.30062079429626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.3190783977508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.33262081146240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.3403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.3464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.3694015979766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.37345280647277834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.37429759502410886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.3914367914199829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.40697598457336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.4102335929870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.42542080879211425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.4743680000305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.4790783882141113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.2,0.5467391967773437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.2,0.6326399803161621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.2,0.750809621810913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.2,0.7674560070037841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.2,0.9188032150268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.2,1.0520511627197267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.2,1.3678848266601562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.2,1.6945024490356446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,balanced,0.07568533221880595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,balanced,0.0805920014778773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,balanced,0.09526933232943217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,balanced,0.13209066788355509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,balanced,0.19806400934855142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,balanced,0.30558933814366657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,balanced,0.30854399998982746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,balanced,0.3121333320935567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,balanced,0.3132693370183309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,balanced,0.31361067295074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,balanced,0.3161440094312032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,balanced,0.32366400957107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,balanced,0.3244640032450358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,balanced,0.332202672958374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,balanced,0.3389813502629598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,balanced,0.344650665918986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,balanced,0.3605653444925944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,balanced,0.3978346586227417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.42794132232666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.506384015083313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.5689440170923868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.7225813070933024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.8765973250071207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,balanced,1.2642186482747395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,balanced,1.5601760546366374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,balanced,2.260810693105062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,balanced,2.9806718826293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,power_law_1.01,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,power_law_1.01,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,power_law_1.01,0.08018559813499451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,power_law_1.01,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,power_law_1.01,0.11562880277633666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,power_law_1.01,0.11870080232620239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,power_law_1.01,0.1236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,power_law_1.01,0.1273408055305481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,power_law_1.01,0.1292672038078308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,power_law_1.01,0.13530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,power_law_1.01,0.14385279417037963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,power_law_1.01,0.1473088026046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,power_law_1.01,0.14777599573135375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,power_law_1.01,0.15684479475021362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,power_law_1.01,0.1622015953063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,power_law_1.01,0.17270400524139404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,power_law_1.01,0.2003648042678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,power_law_1.01,0.22186880111694335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,power_law_1.01,0.27694718837738036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,power_law_1.01,0.32934401035308836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,power_law_1.01,0.4178368091583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,power_law_1.01,0.5650879859924316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,power_law_1.01,0.770739221572876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,power_law_1.01,0.977996826171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,power_law_1.01,1.348902416229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,power_law_1.01,1.7336063385009766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,power_law_1.01,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,power_law_1.01,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,power_law_1.01,0.04204800128936768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,power_law_1.01,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,power_law_1.01,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,power_law_1.01,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,power_law_1.01,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,power_law_1.01,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,power_law_1.01,0.09894400238990783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,power_law_1.01,0.12721920013427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,power_law_1.01,0.16593919992446898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,power_law_1.01,0.20433919429779052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,power_law_1.01,0.27111039161682127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,power_law_1.01,0.36269440650939944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,power_law_1.01,0.49297919273376467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,power_law_1.01,0.6432767868041992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,power_law_1.01,0.06510080099105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,power_law_1.01,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,power_law_1.01,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,power_law_1.01,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,power_law_1.01,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,power_law_1.01,0.0630016028881073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,power_law_1.01,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,power_law_1.01,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,power_law_1.01,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,power_law_1.01,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,power_law_1.01,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,power_law_1.01,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,power_law_1.01,0.10476800203323364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,power_law_1.01,0.13024640083312988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,power_law_1.01,0.14658559560775758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,power_law_1.01,0.18468480110168456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,power_law_1.01,0.22485759258270263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,power_law_1.01,0.3136512041091919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,power_law_1.01,0.4084159851074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,power_law_1.01,0.5644224166870118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,power_law_1.01,0.7250751972198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,balanced,0.052784000833829246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,balanced,0.054287999868392944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,balanced,0.07838400204976399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,balanced,0.09431466460227966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,balanced,0.10826133688290913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,balanced,0.10963732997576396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,balanced,0.10937066872914632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,balanced,0.11033067107200623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,balanced,0.11149866382280986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,balanced,0.11268267035484314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,balanced,0.11386666695276897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,balanced,0.11617599924405415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,balanced,0.11517866452534993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,balanced,0.12166933218638103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,balanced,0.12333866953849792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,balanced,0.12997333208719888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,balanced,0.15918399890263876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,balanced,0.15531733632087708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,balanced,0.19795199235280356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,balanced,0.21477866172790527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,balanced,0.2991146643956502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,balanced,0.2956906755765279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,balanced,0.42694934209187824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,balanced,0.4737279812494914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,balanced,0.6728373368581136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,balanced,0.8156586488087972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,balanced,0.060773332913716636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,balanced,0.0897920032342275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,balanced,0.09245866537094116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,balanced,0.09309867024421692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,balanced,0.09311999877293904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,balanced,0.09262399872144063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,balanced,0.0956106682618459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,balanced,0.09874666730562846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,balanced,0.10802132884661357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,balanced,0.10925333698590596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,balanced,0.11193600296974182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,balanced,0.12594133615493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,balanced,0.13496533036231995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,balanced,0.15716266632080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,balanced,0.1775839924812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,balanced,0.24113066991170248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,balanced,0.27586666742960614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,balanced,0.399893323580424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,balanced,0.49667199452718097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,balanced,0.7128640015920004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,balanced,0.9076639811197916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,balanced,0.05442666510740916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,balanced,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,balanced,0.08120533327261607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,balanced,0.08266666531562805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,balanced,0.08517866333325703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,balanced,0.08444799979527791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,balanced,0.08476266264915466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,balanced,0.08699199557304382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,balanced,0.0928000013033549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,balanced,0.09451199571291606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,balanced,0.10058133800824483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,balanced,0.10717866818110149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,balanced,0.1216319998105367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,balanced,0.1325813333193461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,balanced,0.15981333454449972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,balanced,0.18305599689483643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,balanced,0.24809066454569498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,balanced,0.30452799797058105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,balanced,0.44044268131256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,balanced,0.5721919933954874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,balanced,0.04409599800904592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,balanced,0.04348266621430715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,balanced,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,balanced,0.04614399870236715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,balanced,0.04955733319123586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,balanced,0.06452799836794536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,balanced,0.07255466779073079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,balanced,0.08053866525491078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,balanced,0.09880533814430237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,balanced,0.12146666646003723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,balanced,0.17222400506337485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,balanced,0.20957867304484049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,balanced,0.2917120059331258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,balanced,0.3675466775894165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,power_law_1.2,0.048607999086380006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,power_law_1.2,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,power_law_1.2,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,power_law_1.2,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,power_law_1.2,0.09784319996833801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.2,0.1115455985069275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.2,0.13441920280456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.2,0.16332800388336183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.2,0.21338241100311278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.2,0.26975359916687014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.2,0.4161344051361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.2,0.5863488197326661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.2,0.8704832077026368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.2,1.0568191528320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.17765120267868043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.27808640003204343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.31968638896942136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.4195136070251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.6179391860961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,power_law_1.01,0.8840255737304688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,power_law_1.01,0.9423680305480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,power_law_1.01,0.9605824470520019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,power_law_1.01,0.994598388671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,power_law_1.01,0.9889408111572265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,power_law_1.01,1.0356032371520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.0465855598449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.0653440475463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.0602239608764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.1151424407958985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.125932788848877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.1913344383239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.2616127967834472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,power_law_1.01,1.364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,power_law_1.01,1.5710847854614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,power_law_1.01,1.6832767486572267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,power_law_1.01,1.9246912002563477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,power_law_1.01,2.0624767303466798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,power_law_1.01,2.479443168640137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,power_law_1.01,3.128198432922363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,power_law_1.01,3.9828929901123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,power_law_1.01,5.0469505310058596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,power_law_1.01,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,power_law_1.01,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,power_law_1.01,0.09975039958953857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,power_law_1.01,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,power_law_1.01,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,power_law_1.01,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,power_law_1.01,0.10192639827728271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,power_law_1.01,0.10369280576705933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,power_law_1.01,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,power_law_1.01,0.10941439867019653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,power_law_1.01,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,power_law_1.01,0.11507840156555176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,power_law_1.01,0.12217600345611572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,power_law_1.01,0.13056000471115112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.01,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.01,0.18046720027923585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.01,0.21660799980163575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.01,0.2867072105407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.01,0.35658879280090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.01,0.508358383178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.01,0.6850175857543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.01,1.0397248268127441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.01,1.2537599563598634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.1505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.2024319887161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.20776960849761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.2093951940536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.2198784112930298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.21960959434509278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.22459518909454346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.23129599094390868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.23045759201049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.254035210609436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.25038719177246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.27871999740600584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.27363200187683107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.31432960033416746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.28718719482421873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.34399359226226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,power_law_1.2,0.4055168151855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,power_law_1.2,0.5268479824066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,power_law_1.2,0.5417088031768799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,power_law_1.2,0.7846208095550538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,power_law_1.2,1.007852840423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,power_law_1.2,1.4268480300903321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,power_law_1.2,1.6325056076049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,power_law_1.01,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,power_law_1.01,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,power_law_1.01,0.13751039505004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,power_law_1.01,0.17331839799880983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,power_law_1.01,0.18159359693527222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,power_law_1.01,0.1871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,power_law_1.01,0.1921023964881897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,power_law_1.01,0.2024768114089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,power_law_1.01,0.21585280895233155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,power_law_1.01,0.2224191904067993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,power_law_1.01,0.23872640132904052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,power_law_1.01,0.2585664033889771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,power_law_1.01,0.2705024003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,power_law_1.01,0.26472959518432615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,power_law_1.01,0.3159359931945801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,power_law_1.01,0.3525248050689697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,power_law_1.01,0.4468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,power_law_1.01,0.5271359920501709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,power_law_1.01,0.6922624111175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,power_law_1.01,0.8666367530822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,power_law_1.01,1.1791744232177734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,power_law_1.01,1.5037759780883788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,power_law_1.01,2.1383167266845704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,power_law_1.01,2.809996795654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,power_law_1.2,2.730777549743652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,power_law_1.2,3.7498752593994142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.02919679880142212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.02908160090446472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.029100799560546876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.028870400786399842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.05013759732246399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.09823359847068787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.1332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.1603968024253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.24907519817352294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,power_law_1.2,0.349785590171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.01,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.01,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.01,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.01,0.15700479745864868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.01,0.2170624017715454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.01,0.2305919885635376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.01,0.2247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.01,0.2307136058807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.01,0.23525118827819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.01,0.24079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.01,0.2527551889419556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.01,0.25866239070892333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.01,0.26435840129852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.01,0.28435840606689455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.01,0.2901632070541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.01,0.30371201038360596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.01,0.355948805809021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.01,0.38380160331726076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.01,0.47952637672424314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.01,0.5835264205932618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.01,0.7562367916107178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.01,0.8754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.01,1.1831487655639648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.01,1.4011136054992677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.01,2.0267520904541017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.01,2.64898567199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.0499455988407135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.08936319947242737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.09886720180511474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.10631040334701539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.10898560285568237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.11406079530715943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.12455040216445923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.12811520099639892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.13753600120544435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.16312960386276246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.18751360177993776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.23848960399627686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.28815360069274903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.39587199687957764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.4992256164550781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7083968162536621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9163071632385253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.3237504005432128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.2,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.2,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.2,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.2,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.730860710144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.2,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.2,0.13237119913101197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.2,0.14263039827346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.2,0.1438912034034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.2,0.14565119743347169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.2,0.14819200038909913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.2,0.15319039821624755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.2,0.15557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.2,0.1611199975013733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.2,0.169977605342865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.2,0.17968000173568727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.2,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.2,0.19457919597625734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.2,0.22775681018829347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.2,0.25163519382476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.2,0.30807039737701414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.2,0.37854719161987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.2,0.4839807987213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.2,0.5512127876281738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.2,0.7920512199401856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.2,0.9178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.2,1.2951295852661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.2,1.678656005859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,power_law_1.2,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,power_law_1.2,0.0883903980255127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,power_law_1.2,0.1024448037147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,power_law_1.2,0.10432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,power_law_1.2,0.12798719406127929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,power_law_1.2,0.12889599800109863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,power_law_1.2,0.1327936053276062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,power_law_1.2,0.1317952036857605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,power_law_1.2,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,power_law_1.2,0.13329919576644897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,power_law_1.2,0.14073599576950074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,power_law_1.2,0.14222079515457153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,power_law_1.2,0.1521407961845398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,power_law_1.2,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,power_law_1.2,0.1626431941986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,power_law_1.2,0.17926399707794188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.2,0.19629440307617188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.2,0.23047680854797364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.2,0.2561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.2,0.3147648096084595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.2,0.36918399333953855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.2,0.4961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.2,0.601036787033081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.2,0.8592512130737304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.2,1.1896832466125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.07140480279922486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.08028159737586975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.13056000471115112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.1619647979736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.2124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.24684159755706786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.3578304052352905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.40129919052124025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.5611711978912354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,0.8094655990600585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.1133376121520997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,1.7628608703613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,balanced,0.0544106662273407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,balanced,0.07866666714350383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,balanced,0.11896533767382304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,balanced,0.20491733153661093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,balanced,0.20665067434310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,balanced,0.2074293295542399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,balanced,0.20828266938527426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,balanced,0.20878932873408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,balanced,0.2102186679840088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,balanced,0.2148426572481791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,balanced,0.21758399407068887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,balanced,0.21871467431386313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,balanced,0.2225386699040731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,balanced,0.22806400060653687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,balanced,0.23164800802866617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,balanced,0.2405386765797933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,balanced,0.2560426592826843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,balanced,0.2825439969698588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,balanced,0.316810667514801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,balanced,0.3490080038706462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,balanced,0.43702932198842365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,balanced,0.48923734823862713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,balanced,0.6798453330993652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,balanced,0.8071733315785726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,balanced,1.136181354522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,balanced,1.4621973037719727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,power_law_1.2,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,power_law_1.2,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,power_law_1.2,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.2,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.2,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.2,0.10990079641342163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.2,0.1477120041847229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.2,0.18032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.2,0.24430720806121825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.2,0.3233920097351074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,power_law_1.2,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,power_law_1.2,0.0591808021068573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,power_law_1.2,0.060108798742294314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,power_law_1.2,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,power_law_1.2,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,power_law_1.2,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,power_law_1.2,0.10124160051345825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,power_law_1.2,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,power_law_1.2,0.13371520042419432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,power_law_1.2,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,power_law_1.2,0.21054720878601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,power_law_1.2,0.28640639781951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,power_law_1.2,0.35921919345855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,power_law_1.2,0.4873472213745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,power_law_1.2,0.6340479850769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,balanced,0.1933280030886332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,balanced,0.27659199635187787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,balanced,0.44976532459259033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,balanced,0.795792023340861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,balanced,1.1406559944152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,balanced,1.1488479773203533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,balanced,1.1572106679280598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,balanced,1.1617759863535564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,balanced,1.168554703394572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,balanced,1.1779306729634602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,balanced,1.190618673960368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,balanced,1.2004533608754475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,balanced,1.2164426644643147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,balanced,1.242410659790039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,balanced,1.2677546342213948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,balanced,1.305349349975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,balanced,1.3991306622823079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,balanced,1.485322634379069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,balanced,2.1172107060750327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,balanced,1.8637226422627766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,balanced,2.6310240427652993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,balanced,2.7853867212931314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,balanced,4.147237459818522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,balanced,4.675050735473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,balanced,7.188085556030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,balanced,8.481818517049154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,balanced,0.05425600210825602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,balanced,0.04865066707134247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,balanced,0.055685331424077354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,balanced,0.05763733386993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,balanced,0.06380266447861989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.09116799632708232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.11155733466148376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.1328053375085195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.1455413301785787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.18480533361434937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.22087466716766357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.3031253417332967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.38310400644938153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.2,0.16654080152511597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.2,0.23311359882354737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.2,0.32632319927215575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.2,0.40327038764953616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.2,0.6710656166076661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.2,0.6899456024169922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.2,0.7071487903594971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.2,0.724012804031372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.2,0.7367040157318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.2,0.7588352203369141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.2,0.7791488170623779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.2,0.8094400405883789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.2,0.8215807914733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.2,0.8622783660888672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.2,0.8844160079956055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.2,0.9155839920043946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.2,1.019007968902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.2,1.0972415924072265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.2,1.2990336418151855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.2,1.3165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.2,1.6052799224853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.2,1.6828927993774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.2,2.10067195892334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.2,2.5466047286987306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.2,3.6469566345214846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.2,4.119891357421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,balanced,0.0399893323580424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,balanced,0.04083733260631561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,balanced,0.04135466615358988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,balanced,0.04073066761096319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,balanced,0.04618666569391886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,balanced,0.048298666874567665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,balanced,0.06461866696675618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,balanced,0.08827199538548787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,balanced,0.10271466771761577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,balanced,0.057616000374158226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,balanced,0.06251200040181477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,balanced,0.06444266438484192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,balanced,0.0637546678384145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,balanced,0.07051733136177063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,balanced,0.07075733443101247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,balanced,0.07329600056012471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,balanced,0.08899733424186707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,balanced,0.08145066599051158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,balanced,0.0944053332010905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,balanced,0.09850133458773296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,balanced,0.10313066840171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,balanced,0.12763200203577676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,balanced,0.13877333203951517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,balanced,0.19433599710464478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,balanced,0.233952005704244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,balanced,0.3327839970588684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,balanced,0.4089653491973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,balanced,0.6174079974492391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,balanced,0.7818026542663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.15782400369644164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.1909183979034424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.18789759874343873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.1889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.19114880561828612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.19396480321884155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.20072319507598876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.19997440576553344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.20714240074157714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.21305599212646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.2218559980392456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.22903039455413818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,power_law_1.2,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.2280639886856079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,power_law_1.2,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.24906880855560304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.2640255928039551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,power_law_1.2,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.29818880558013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.3171135902404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,power_law_1.2,0.15665279626846312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.3969088077545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,power_law_1.2,0.16213760375976563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,power_law_1.01,0.44618239402771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,power_law_1.2,0.1644863963127136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,power_law_1.01,0.5523263931274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,power_law_1.01,0.6427711963653564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,power_law_1.01,1.1140480041503906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,power_law_1.01,1.1347455978393555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,power_law_1.2,0.17612799406051635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,power_law_1.2,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,power_law_1.2,0.18540799617767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,power_law_1.2,0.18876800537109376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,power_law_1.2,0.19251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,power_law_1.2,0.2020672082901001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,power_law_1.2,0.21335039138793946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,power_law_1.2,0.22042880058288575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,power_law_1.2,0.2368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,power_law_1.2,0.27457919120788576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,power_law_1.2,0.30319359302520754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,power_law_1.2,0.36716160774230955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,power_law_1.2,0.4358975887298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,power_law_1.2,0.5776512145996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,power_law_1.2,0.6540863990783692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,power_law_1.2,0.8923199653625489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,power_law_1.2,1.0595647811889648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,power_law_1.2,1.495199966430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,power_law_1.2,1.897235107421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,balanced,0.029861333469549816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,balanced,0.030965333183606465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,balanced,0.03130666663249334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,balanced,0.03388266762097677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,balanced,0.032101333141326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,balanced,0.03514666606982549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,balanced,0.03384533276160558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,balanced,0.03643733263015747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,balanced,0.03723733375469843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,balanced,0.048138668139775596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,balanced,0.057205334305763245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,balanced,0.06136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,balanced,0.07868266602357228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,balanced,0.09810133775075276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,balanced,0.13621333241462708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,balanced,0.15609600146611533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,balanced,0.2071359952290853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,balanced,0.2441706657409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,balanced,0.3557973305384318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,balanced,0.46133867899576825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.01,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.01,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.01,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.01,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.01,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.01,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.01,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.01,0.12327680587768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.01,0.14979840517044068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.01,0.1923840045928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.01,0.2043776035308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.01,0.26727681159973143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.01,0.33574399948120115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.01,0.4490816116333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.01,0.562611198425293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.18435200452804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.22776319980621337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.30587520599365237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.3204416036605835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.3353280067443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.34179201126098635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.34542078971862794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.3639103889465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.373145604133606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.3817471981048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.397324800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.40687999725341795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.42855038642883303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.4874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.52707839012146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.6205887794494629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.6917439937591553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.8791168212890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.0508031845092773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,1.4214464187622071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,1.7791360855102538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,2.4997184753417967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,3.242099380493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,power_law_1.01,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,power_law_1.01,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,power_law_1.01,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,power_law_1.01,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,power_law_1.01,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.01,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.01,0.12250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.01,0.14877439737319947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.01,0.20207359790802001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.01,0.25321600437164304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.01,0.38835198879241944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.0686464011669159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.08228480219841003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.01,0.45980157852172854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.15092480182647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.1820672035217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.24053120613098145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.283027195930481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.4271423816680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,power_law_1.01,0.49015040397644044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.01,0.7760064125061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.14856959581375123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.15190399885177613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.16151039600372313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.16210559606552125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.16229759454727172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.16535040140151977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.16597119569778443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.1712448000907898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.17671040296554566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.18127360343933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18507519960403443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.2076672077178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.20894720554351806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.22939519882202147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.2635200023651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.30687360763549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.3767296075820923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.4600512027740479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.613862419128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.7628608226776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.01,0.9257344245910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.1016639709472655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,power_law_1.01,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,power_law_1.01,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,power_law_1.01,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,power_law_1.01,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,power_law_1.01,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,power_law_1.01,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,power_law_1.01,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,power_law_1.01,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,power_law_1.01,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.01,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.01,0.09765759706497193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.01,0.12295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.01,0.15745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.01,0.18982399702072145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.01,0.2682751893997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.01,0.35920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.01,0.5690879821777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.01,0.7782015800476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.4308927536010743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,2.1537664413452147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,2.6533376693725588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.2,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.2,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.2,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.2,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.2,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.2,0.05538560152053833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.2,0.061375999450683595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.2,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.2,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.2,0.13984639644622804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.2,0.15448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.2,0.2412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.2,0.32549118995666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.15831040143966674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.20653440952301025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.21376640796661378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.22152318954467773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.2285696029663086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.24112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.23898239135742189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.2528127908706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.26384639739990234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.26988160610198975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.2802047967910767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.30462079048156737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.3250240087509155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.3831615924835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.44435839653015136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.5523263931274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,power_law_1.01,0.6494912147521973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,power_law_1.01,0.8762240409851074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.0664511680603028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,power_law_1.01,1.5344639778137208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,power_law_1.01,1.726041603088379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,balanced,0.036176001032193504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,balanced,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,balanced,0.04246933261553446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,balanced,0.0631573349237442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,balanced,0.08561066786448161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,balanced,0.09983999530474345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,balanced,0.13544533650080362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,balanced,0.16887466112772623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,balanced,0.22844799359639487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,balanced,0.2847946683565776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,power_law_1.2,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,power_law_1.2,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,power_law_1.2,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,power_law_1.2,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,power_law_1.2,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,power_law_1.2,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,power_law_1.2,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,power_law_1.2,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,power_law_1.2,0.07820159792900086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,power_law_1.2,0.11908479928970336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,power_law_1.2,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,power_law_1.2,0.20944640636444092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,power_law_1.2,0.26695680618286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,power_law_1.2,0.3939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,power_law_1.2,0.4787007808685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.06981760263442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.08015999794006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.09359359741210938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.12755839824676513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.12849279642105102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.17093119621276856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.18561919927597045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.2708672046661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.331494402885437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.4400063991546631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.5991168022155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,power_law_1.2,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,power_law_1.2,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,power_law_1.2,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,power_law_1.2,0.11100159883499146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,power_law_1.2,0.1118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,power_law_1.2,0.1295807957649231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,power_law_1.2,0.13404799699783326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,power_law_1.2,0.13286399841308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,power_law_1.2,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,power_law_1.2,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,power_law_1.2,0.14174720048904418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,power_law_1.2,0.14581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,power_law_1.2,0.14974720478057862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,power_law_1.2,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,0.7817599773406982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,power_law_1.2,0.15556479692459108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,power_law_1.2,0.15944960117340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,power_law_1.2,0.17533440589904786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.2,0.19196159839630128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.2,0.21580801010131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.2,0.251692795753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.2,0.3106879949569702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.2,0.3892287969589233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.2,0.5367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.2,0.6901696205139161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.2,0.9664575576782226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.2,1.2724224090576173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.082374382019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,power_law_1.01,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,power_law_1.01,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,power_law_1.01,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,power_law_1.01,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,power_law_1.01,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,power_law_1.01,0.09004799723625183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,power_law_1.01,0.10732159614562989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,power_law_1.01,0.12681599855422973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,power_law_1.01,0.1592576026916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,power_law_1.01,0.21288959980010985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,power_law_1.01,0.2978431940078735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,power_law_1.01,0.3424704074859619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,power_law_1.01,0.5520639896392823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,power_law_1.01,0.7230783939361572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,balanced,0.047498668233553566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,balanced,0.04961599906285604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,balanced,0.05349866549173991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,balanced,0.06409599880377452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,balanced,0.06593066453933716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,balanced,0.09554133812586467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,balanced,0.1113813320795695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,balanced,0.15014400084813437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,balanced,0.17828800280888876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,balanced,0.24262932936350504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,balanced,0.3033386667569478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.060121601819992064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.06984959840774536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.0882752001285553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.0972927987575531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.12149120569229126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.14327679872512816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.19252480268478395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.21488640308380128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.2965759992599487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.3720128059387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.5570432186126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.6607935905456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.0986240029335022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.14706560373306274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.2008768081665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.2848448038101196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.38906240463256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.4689343929290771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.6065023899078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.6712384223937988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.6685376167297363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.6565375804901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.7217472076416016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.7484032154083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.7262271881103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.766649580001831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,power_law_1.01,0.7585984230041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,power_law_1.01,0.7889535903930665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,power_law_1.01,0.819974422454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,power_law_1.01,0.8885120391845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,power_law_1.01,0.8979519844055176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,power_law_1.01,0.9742976188659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.0229248046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,power_law_1.01,1.1881664276123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,power_law_1.01,1.2971455574035644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,power_law_1.01,1.6547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,power_law_1.01,1.8906112670898438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,power_law_1.01,2.413420867919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,power_law_1.01,2.9330368041992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,power_law_1.2,0.02565760016441345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,power_law_1.2,0.02455040067434311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,power_law_1.2,0.02622720003128052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,power_law_1.2,0.027142399549484254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,power_law_1.2,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,power_law_1.2,0.02927359938621521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,power_law_1.2,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,power_law_1.2,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,power_law_1.2,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,power_law_1.2,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,power_law_1.2,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,power_law_1.2,0.12965760231018067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,power_law_1.2,0.2627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,power_law_1.2,0.3747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,balanced,0.029717333614826202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,balanced,0.03935466706752777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,balanced,0.03878399978081385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,balanced,0.05783999959627787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,balanced,0.08178666730721791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,balanced,0.11098133524258931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,balanced,0.1386613349119822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,balanced,0.1916266679763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,balanced,0.23221866289774576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,balanced,0.3290666739145915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,balanced,0.4257066647211711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,power_law_1.2,0.5235455989837646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,power_law_1.2,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,power_law_1.2,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,power_law_1.2,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,power_law_1.2,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,power_law_1.2,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,power_law_1.2,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,power_law_1.2,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,power_law_1.2,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.2,0.12750719785690307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.2,0.1559872031211853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.2,0.2125632047653198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.2,0.278656005859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,power_law_1.2,0.7271935939788818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,balanced,0.07006399830182393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,balanced,0.09700266520182292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,balanced,0.1343893309434255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.2,0.3757375955581665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,balanced,0.21902932723363241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,balanced,0.34815998872121173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,balanced,0.4453866481781006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,balanced,0.4458826780319214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,balanced,0.4497493505477905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,balanced,0.4514186779658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,balanced,0.4522879918416341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,balanced,0.4529706637064616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,balanced,0.45604264736175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,balanced,0.45524799823760986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,balanced,0.4604479869206746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,balanced,0.4663360118865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,balanced,0.46902934710184735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,balanced,0.47679467995961505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,balanced,0.525546669960022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,balanced,0.5083200136820475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,balanced,0.6334826548894247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,balanced,0.5760853290557861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,balanced,0.9521973133087158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,balanced,0.6935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,balanced,1.0876959959665935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,balanced,1.0083306630452473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,balanced,1.6560959815979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,balanced,1.5250879923502605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.2,0.5854527950286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.2,0.9525952339172363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.2,1.3023551940917968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,power_law_1.2,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,power_law_1.2,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,power_law_1.2,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,power_law_1.2,0.12081279754638671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,power_law_1.2,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,power_law_1.2,0.13100160360336305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,power_law_1.2,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,power_law_1.2,0.15203839540481567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,power_law_1.2,0.14713599681854247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,power_law_1.2,0.1576192021369934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,power_law_1.2,0.16278400421142578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,power_law_1.2,0.1757248044013977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,power_law_1.2,0.20051839351654052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,power_law_1.2,0.23578240871429443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,power_law_1.2,0.2789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,power_law_1.2,0.3346368074417114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,power_law_1.2,0.43632001876831056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,power_law_1.2,0.5676991939544678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.11425919532775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.1681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.16394879817962646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.18830080032348634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.22108800411224366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,power_law_1.2,0.8105855941772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.23419520854949952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.20563199520111083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.21778559684753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.21619839668273927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.19225599765777587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.21264638900756835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.20755200386047362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.2178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.23125760555267333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.21899518966674805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.24543359279632568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,power_law_1.2,1.0243071556091308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.26502399444580077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.2690943956375122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.3033663988113403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.33665919303894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.4211711883544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.4650047779083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.6014976024627685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,0.718009614944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,0.9847231864929199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.2880895614624024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,power_law_1.2,1.4358207702636718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,power_law_1.2,1.855936050415039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,balanced,0.060640002290407814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,balanced,0.08161066472530365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,balanced,0.08452266454696655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,balanced,0.08540266752243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,balanced,0.08679466446240743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,balanced,0.0867680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,balanced,0.08733866612116496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,balanced,0.09417600433031718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,balanced,0.09357333183288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,balanced,0.09648000200589497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,balanced,0.10501866539319356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.10776533683141072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.1211946705977122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.13570666313171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.15731733043988547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.18037333091100058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.23160000642140707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.2752053340276082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.37670401732126874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.46080533663431805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,power_law_1.2,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,power_law_1.2,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,power_law_1.2,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,power_law_1.2,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,power_law_1.2,0.10294400453567505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,power_law_1.2,0.11045119762420655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,power_law_1.2,0.11501439809799194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,power_law_1.2,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,power_law_1.2,0.13946880102157594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,power_law_1.2,0.17036800384521483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,power_law_1.2,0.20958080291748046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,power_law_1.2,0.2736191987991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,power_law_1.2,0.3748928070068359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,power_law_1.2,0.453542423248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,power_law_1.2,0.5532032012939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,power_law_1.2,0.8064191818237305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,power_law_1.2,1.0430015563964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,power_law_1.2,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,power_law_1.2,0.07301120162010193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,power_law_1.2,0.0759935975074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,power_law_1.2,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,power_law_1.2,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,power_law_1.2,0.11411839723587036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,power_law_1.2,0.13597439527511596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,power_law_1.2,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,power_law_1.2,0.20992639064788818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,power_law_1.2,0.267193603515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,power_law_1.2,0.36212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,power_law_1.2,0.47285118103027346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,power_law_1.2,0.7221695899963378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.1065343976020813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.12513920068740844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.19811840057373048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.1958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.2043328046798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.20675199031829833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.20773119926452638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.21480960845947267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.21420159339904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.22702720165252685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.22990078926086427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.23308799266815186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.2529792070388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.2727231979370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.2671360015869141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.3145472049713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,power_law_1.2,0.8829824447631835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.33889920711517335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.41317119598388674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,power_law_1.2,0.48438401222229005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,power_law_1.2,0.6369791984558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,power_law_1.2,0.7955840110778809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,power_law_1.2,1.0807680130004882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,power_law_1.2,1.423417568206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,power_law_1.2,2.027507209777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,power_law_1.2,2.6887168884277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,balanced,0.046096002062161766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,balanced,0.045514668027559914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,balanced,0.04738666613896688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,balanced,0.04667733112970988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,balanced,0.047413334250450134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,balanced,0.048623998959859215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,balanced,0.04645333190759023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,balanced,0.04836800197760264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,balanced,0.05152533451716105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,balanced,0.06266133487224579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,balanced,0.08258666594823201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,balanced,0.09519466757774353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,balanced,0.11917333801587422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,balanced,0.15005866686503092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,balanced,0.027600000301996868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,balanced,0.029946667452653248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,balanced,0.047584002216657005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,balanced,0.05097066859404246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,balanced,0.1055519978205363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,balanced,0.1379093329111735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,balanced,0.1827146609624227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,balanced,0.22381333510080972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,balanced,0.3192906578381856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,balanced,0.4061973492304484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,power_law_1.2,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,power_law_1.2,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,power_law_1.2,0.05059840083122254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,power_law_1.2,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,power_law_1.2,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,power_law_1.2,0.0573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,power_law_1.2,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,power_law_1.2,0.08175359964370728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,power_law_1.2,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,power_law_1.2,0.12161920070648194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,power_law_1.2,0.14850560426712037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,power_law_1.2,0.1965440034866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,power_law_1.2,0.24130558967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,power_law_1.2,0.3505856037139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,power_law_1.2,0.4083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,power_law_1.2,0.6777088165283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,power_law_1.2,0.8477567672729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,power_law_1.2,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,power_law_1.2,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,power_law_1.2,0.0719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,power_law_1.2,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,power_law_1.2,0.08555520176887513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,power_law_1.2,0.08574079871177673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,power_law_1.2,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,power_law_1.2,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,power_law_1.2,0.10652799606323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,power_law_1.2,0.11141120195388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,power_law_1.2,0.11758079528808593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,power_law_1.2,0.12878719568252564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,power_law_1.2,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,power_law_1.2,0.17383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,power_law_1.2,0.2106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.2,0.2515007972717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.2,0.31807360649108884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.2,0.42048001289367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.2,0.5489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.2,0.7698815822601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.2,1.1576895713806152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.2,1.4325695991516114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.2,1.8454719543457032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,balanced,0.04937066634496053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06624533236026764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,balanced,0.0680159976085027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,balanced,0.07100266714890797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,balanced,0.0773226668437322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.09835199515024821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.13402666648228964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.15761599938074747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.21222933133443198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.2454506754875183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.3431679805119832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.44943467775980633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.2,3.0396928787231445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,power_law_1.2,0.0749504029750824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,power_law_1.2,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,power_law_1.2,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,power_law_1.2,0.10469759702682495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,power_law_1.2,0.10312960147857667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,power_law_1.2,0.10577919483184814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,power_law_1.2,0.1076416015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,power_law_1.2,0.10616320371627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,power_law_1.2,0.1059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,power_law_1.2,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,power_law_1.2,0.1118656039237976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,power_law_1.2,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,power_law_1.2,0.12131199836730958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,power_law_1.2,0.12568960189819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,power_law_1.2,0.13138560056686402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,power_law_1.2,0.14505599737167357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.2,0.15783679485321045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.2,0.19127680063247682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.2,0.21933441162109374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.2,0.2935424089431763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.2,0.3785792112350464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.2,0.5298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.2,0.7051583766937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.2,1.02805757522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.2,1.2469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.01,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.01,0.10776319503784179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.01,0.11163519620895386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.01,0.11861120462417603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.01,0.12903679609298707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.01,0.12536319494247436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.01,0.1285055994987488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.01,0.12776960134506227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.01,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.01,0.13111040592193604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.01,0.13516160249710082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.01,0.13919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.01,0.14464000463485718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.01,0.15159679651260377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.01,0.15779199600219726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.01,0.16662399768829345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.01,0.18981759548187255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.01,0.21854720115661622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.01,0.27314560413360595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.01,0.3101056098937988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.01,0.4087103843688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.01,0.46136960983276365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.01,0.6084352016448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.01,0.7608704090118408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.01,1.1161215782165528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.01,1.4689727783203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,balanced,0.04196799794832865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,balanced,0.04123199979464213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,balanced,0.05228800078233083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,balanced,0.054986665646235146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,balanced,0.05576533575852712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,balanced,0.05563733478387197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,balanced,0.05952533086140951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,balanced,0.07796800136566162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,balanced,0.0892693301041921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,balanced,0.09504000345865886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,balanced,0.11387733618418376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,balanced,0.13054933150609335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,balanced,0.1625759998957316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,balanced,0.19607466459274292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,balanced,0.2606400052706401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,balanced,0.32546667257944745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,balanced,0.46722666422526044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,balanced,0.5816053152084351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,balanced,0.04919999837875366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,balanced,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,balanced,0.06881066660086314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,balanced,0.06932266553243001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,balanced,0.07142933209737141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,balanced,0.07242666681607564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,balanced,0.07658666869004567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,balanced,0.08127466837565105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,balanced,0.08353066444396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,balanced,0.09115733702977498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,balanced,0.09966933727264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,balanced,0.1064906617005666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,balanced,0.12573867042859396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,balanced,0.1442346672217051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,balanced,0.17906665802001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,balanced,0.2150826652844747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,balanced,0.28489067157109577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,balanced,0.3521546522776286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,balanced,0.48852264881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,balanced,0.6284533341725668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,power_law_1.2,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,power_law_1.2,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,power_law_1.2,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,power_law_1.2,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,power_law_1.2,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,power_law_1.2,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,power_law_1.2,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,power_law_1.2,0.11015039682388306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,power_law_1.2,0.14581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,power_law_1.2,0.21154561042785644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,power_law_1.2,0.2707520008087158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,power_law_1.2,0.395251202583313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,balanced,0.06644266843795776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,balanced,0.0823413332303365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,balanced,0.08244266609350841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,balanced,0.08249066770076752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,balanced,0.08261333405971527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,balanced,0.08582933743794759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,balanced,0.088837335507075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,balanced,0.08806399504343669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,balanced,0.08847999572753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,balanced,0.09505599737167358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,balanced,0.09687999884287517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,balanced,0.10027733445167542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,balanced,0.11426132917404175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,balanced,0.11985066533088684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,balanced,0.14383467038472494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,balanced,0.1686240037282308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,balanced,0.22003199656804404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,balanced,0.21222400665283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,balanced,0.28149332602818805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,balanced,0.32305065790812176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,balanced,0.4487680196762085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,balanced,0.5380586783091227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,power_law_1.2,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,power_law_1.2,0.1079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,power_law_1.2,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,power_law_1.2,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,power_law_1.2,0.09783040285110474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,power_law_1.2,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,power_law_1.2,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,power_law_1.2,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,power_law_1.2,0.10291199684143067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,power_law_1.2,0.10401920080184937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,power_law_1.2,0.11562880277633666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,power_law_1.2,0.11653120517730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,power_law_1.2,0.125273597240448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,power_law_1.2,0.14044159650802612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,power_law_1.2,0.1487104058265686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,power_law_1.2,0.1715775966644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,power_law_1.2,0.18551679849624633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,power_law_1.2,0.22437760829925538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.06501759886741638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,power_law_1.2,0.2313152074813843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,power_law_1.2,0.3293567895889282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.07570559978485107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,power_law_1.2,0.40823678970336913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.09374719858169556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,power_law_1.2,0.5755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.1286784052848816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,power_law_1.2,0.6722879886627198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.21638400554656984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.24276480674743653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,power_law_1.01,0.3715903997421265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,power_law_1.01,0.39680640697479247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,power_law_1.2,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,power_law_1.2,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,power_law_1.2,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,power_law_1.2,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,power_law_1.2,0.08640000224113464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,power_law_1.2,0.08912640213966369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,power_law_1.2,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,power_law_1.2,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,power_law_1.2,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,power_law_1.2,0.0939520001411438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,power_law_1.2,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,power_law_1.2,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,power_law_1.2,0.11592960357666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,power_law_1.2,0.12091519832611083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,power_law_1.2,0.1391487956047058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.2,0.18099199533462523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.2,0.20784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.2,0.27302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.2,0.3581696033477783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.2,0.4733759880065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.2,0.6679296016693115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.2,0.8790528297424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.2,1.2302271842956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,power_law_1.01,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,power_law_1.01,0.054099202156066895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,power_law_1.01,0.06046079993247986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,power_law_1.01,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,power_law_1.01,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,power_law_1.01,0.06702079772949218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,power_law_1.01,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,power_law_1.01,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,power_law_1.01,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,power_law_1.01,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,power_law_1.01,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,power_law_1.01,0.10223360061645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,power_law_1.01,0.12344319820404052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,power_law_1.01,0.14248960018157958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.01,0.1215872049331665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,power_law_1.01,0.1816831946372986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.01,0.14410239458084106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,power_law_1.01,0.2155071973800659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,power_law_1.01,0.29966719150543214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.01,0.17995519638061525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,power_law_1.01,0.38159999847412107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.01,0.21704959869384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,power_law_1.01,0.5446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.01,0.3055488109588623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,power_law_1.01,0.7081344127655029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,power_law_1.01,1.0465920448303223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,power_law_1.01,1.366220760345459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.01,0.3712127923965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.01,0.5828415870666503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.01,0.8008255958557129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,balanced,0.054341331124305725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,balanced,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,balanced,0.07236800094445546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,balanced,0.09860266248385112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,balanced,0.10506666700045268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,balanced,0.10703466335932414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,balanced,0.1093386709690094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,balanced,0.10879466931025188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,balanced,0.10809066891670227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,balanced,0.1113866666952769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,balanced,0.11172800262769063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,balanced,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,balanced,0.11624000469843547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,balanced,0.12430933117866516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,balanced,0.12638933459917703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,balanced,0.13013866543769836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,balanced,0.14361600081125894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.15293866395950317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.1813973387082418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.19928000370661417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.26634132862091064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.29819732904434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.42898134390513104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.5261066754659017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.7597440083821615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.9554879665374756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,balanced,0.08218133449554443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,balanced,0.11963733037312825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,balanced,0.1584160029888153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,balanced,0.19291732708613077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,balanced,0.19391467173894247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,balanced,0.19524800777435303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,balanced,0.1959893306096395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,balanced,0.19580799341201782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,balanced,0.20089600483576456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,balanced,0.20317866404851279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,balanced,0.20620266596476236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,balanced,0.20694933334986368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,balanced,0.21483200788497925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,balanced,0.21577600638071695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,balanced,0.2257173260052999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,balanced,0.24285866816838583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,balanced,0.25699732700983685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,balanced,0.32705599069595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,balanced,0.32867199182510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,balanced,0.468506654103597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,balanced,0.4644800027211507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,balanced,0.6693440278371176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,balanced,0.7351253032684326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,balanced,1.0721226533253987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,balanced,1.2859093348185222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,balanced,0.053082664807637535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,balanced,0.063701331615448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,balanced,0.06543466448783875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,balanced,0.06727999945481618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,balanced,0.06989333530267079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,balanced,0.06963199873765309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,balanced,0.0705813318490982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,balanced,0.0743093341588974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,balanced,0.07730666796366374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,balanced,0.08287466565767924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,balanced,0.08389866352081299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,balanced,0.09310400485992432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,balanced,0.09699733058611552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,balanced,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,balanced,0.1384266714255015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,balanced,0.17612799008687338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,balanced,0.1989226738611857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,balanced,0.2747360070546468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,balanced,0.32710399230321247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,balanced,0.4543199936548869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,balanced,0.5702399810155233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,power_law_1.01,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,power_law_1.01,0.029478400945663452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,power_law_1.01,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,power_law_1.01,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,power_law_1.01,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,power_law_1.01,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,power_law_1.01,0.06609280109405517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,power_law_1.01,0.08954240083694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,power_law_1.01,0.10862079858779908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,power_law_1.01,0.15726079940795898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,power_law_1.01,0.17709439992904663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,power_law_1.01,0.26673920154571534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,power_law_1.01,0.30856320858001707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,power_law_1.01,0.47060480117797854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,power_law_1.01,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,power_law_1.01,0.09656959772109985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,power_law_1.01,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,power_law_1.01,0.13985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,power_law_1.01,0.1433087944984436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,power_law_1.01,0.14703359603881835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,power_law_1.01,0.15235840082168578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,power_law_1.01,0.1588736057281494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,power_law_1.01,0.16844799518585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,power_law_1.01,0.17571200132369996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,power_law_1.01,0.17482240200042726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,power_law_1.01,0.18844799995422362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,power_law_1.01,0.19813120365142822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,power_law_1.01,0.21772799491882325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,power_law_1.01,0.26245760917663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,power_law_1.01,0.30512640476226804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,power_law_1.01,0.38906240463256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,power_law_1.01,0.4611072063446045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,power_law_1.01,0.6152256011962891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,power_law_1.01,0.8035455703735351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,power_law_1.01,1.09486722946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,power_law_1.01,1.4394240379333496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,power_law_1.01,2.0166271209716795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,power_law_1.01,2.583807945251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,power_law_1.01,0.596563196182251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,power_law_1.01,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,power_law_1.01,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,power_law_1.01,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,power_law_1.01,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,power_law_1.01,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,power_law_1.01,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,power_law_1.01,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,power_law_1.01,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.01,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.01,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.01,0.12142080068588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.01,0.14369920492172242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.01,0.18302079439163207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.01,0.2236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.01,0.3167871952056885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.01,0.39112319946289065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.2,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.2,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.2,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.2,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.2,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.2,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.2,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.2,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.2,0.09886720180511474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.2,0.10495359897613525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.2,0.1343616008758545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.2,0.15970560312271118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.2,0.21649279594421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.2,0.26905601024627684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.2,0.3994879961013794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.11413120031356812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.1201856017112732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.12143360376358033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.12765439748764038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.12892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.14929920434951782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.15173120498657228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.15565439462661743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.2,0.4541056156158447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.1789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.1968191981315613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.24810879230499266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.2891648054122925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.3729536056518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,power_law_1.2,0.44663681983947756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,power_law_1.2,0.6276735782623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,power_law_1.2,0.7839168071746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,power_law_1.2,1.1106240272521972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,power_law_1.2,1.406112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.07136639952659607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.09365760087966919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.11924480199813843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.07282559871673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.10423680543899536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.17505919933319092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.12751359939575196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.1324671983718872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.12943359613418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.14152319431304933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.1547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.17359360456466674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.2248703956604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.26158719062805175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.33870720863342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.41300477981567385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.5685760021209717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.7070720195770264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.0106047630310058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.1818176031112671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.3129728317260743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.23111040592193605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,balanced,0.07299200197060902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,balanced,0.09619733691215515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,balanced,0.14852799971898398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,balanced,0.24554665883382162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,balanced,0.4373973210652669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,balanced,0.8333226839701334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,balanced,0.8384106953938802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,balanced,0.8348106543223063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,balanced,0.8386293252309164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,balanced,0.8420320351918539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,balanced,0.8475893338521322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,balanced,0.8586560090382894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,balanced,0.8626399834950765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,balanced,0.8876960277557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,balanced,0.8919146855672201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,balanced,0.899616003036499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,balanced,0.9084479808807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,balanced,0.9565653006235758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,balanced,0.9931626319885254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,balanced,1.0904213587443035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,balanced,1.1384906768798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,balanced,1.2755253314971924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,balanced,1.4231786727905273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,balanced,1.8383839925130208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,balanced,2.070250670115153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,balanced,2.90775458017985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,balanced,3.5605494181315103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.2768127918243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.4713791847229004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.5420735836029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.09904639720916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.10291199684143067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.09450240135192871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.10158720016479492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.10396800041198731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.13210879564285277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.14618879556655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.16694400310516358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.19204479455947876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.24046719074249268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.38202240467071535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.44960641860961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.6431424140930175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,0.7842432022094726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.08311039805412293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.08602240085601806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.08208640217781067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.14878079891204835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.17208319902420044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.21356160640716554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.2585088014602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.34165120124816895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.42466559410095217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,power_law_1.2,0.6172351837158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,power_law_1.2,0.7631999969482421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,power_law_1.01,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,power_law_1.01,0.12650879621505737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,power_law_1.01,0.13489279747009278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,power_law_1.01,0.13505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,power_law_1.01,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,power_law_1.01,0.14440959692001343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,power_law_1.01,0.1506816029548645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,power_law_1.01,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,power_law_1.01,0.153164803981781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,power_law_1.01,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,power_law_1.01,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,power_law_1.01,0.16988799571990967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,power_law_1.01,0.1701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,power_law_1.01,0.18140159845352172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,power_law_1.01,0.19841279983520507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.01,0.2168895959854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.01,0.25878400802612306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.01,0.3068608045578003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.01,0.4139967918395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.01,0.49098877906799315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.01,0.6859392166137696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.01,0.9052543640136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.01,1.2364992141723632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.01,1.5979455947875976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.05066879987716675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.061977601051330565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.14520959854125975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.1646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.22310400009155273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.2732032060623169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.4036096096038818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.4946752071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,balanced,0.029306667546431225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,balanced,0.03382933388153712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,balanced,0.03259733319282532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,balanced,0.033189333975315094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,balanced,0.06647466619809468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,balanced,0.10122666756312053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,power_law_1.2,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,power_law_1.2,0.11391999721527099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,power_law_1.2,0.1382848024368286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,power_law_1.2,0.2259648084640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,power_law_1.2,0.23588480949401855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,power_law_1.2,0.24068479537963866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,power_law_1.2,0.24111359119415282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,power_law_1.2,0.24476161003112792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,power_law_1.2,0.2525696039199829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,power_law_1.2,0.2555775880813599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,power_law_1.2,0.2561856031417847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,power_law_1.2,0.2679231882095337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,power_law_1.2,0.2701503992080688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,power_law_1.2,0.29171841144561766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,power_law_1.2,0.314355206489563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,power_law_1.2,0.3264575958251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,power_law_1.2,0.3356287956237793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,power_law_1.2,0.38779520988464355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,power_law_1.2,0.4382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,power_law_1.2,0.5478464126586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,power_law_1.2,0.6431680202484131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,power_law_1.2,0.8613951683044434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,power_law_1.2,1.0870976448059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,power_law_1.2,1.5115455627441405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,power_law_1.2,1.9516735076904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,power_law_1.2,2.8347776412963865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.06393600006898244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.0803306649128596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.10308266679445903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.12689600388209024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.16053866346677145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.17946134010950723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.24927467107772827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.3156053423881531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,power_law_1.2,3.7247550964355467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,balanced,0.037050666908423104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,balanced,0.037290667494138084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,balanced,0.06458666423956554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,balanced,0.07327466706434886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,balanced,0.11825066804885864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,balanced,0.14050666491190592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,balanced,0.19268266359965006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,balanced,0.24757333596547446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,balanced,0.37088533242543537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,balanced,0.5013973315556844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,power_law_1.2,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,power_law_1.2,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,power_law_1.2,0.062080001831054686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,power_law_1.2,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,power_law_1.2,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,power_law_1.2,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,power_law_1.2,0.08161280155181885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,power_law_1.2,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,power_law_1.2,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.2,0.1091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.2,0.1811776041984558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.2,0.22880640029907226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.2,0.2590208053588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.2,0.4454336166381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.2,0.5416192054748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,balanced,0.03400533397992452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,balanced,0.03467733412981033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,balanced,0.03589866558710734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,balanced,0.03675200045108795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,balanced,0.05163733164469401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,balanced,0.07123200098673503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,balanced,0.07401599983374278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,balanced,0.08875200152397156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,balanced,0.10441600282986958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,balanced,0.1299626628557841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,balanced,0.15902933478355408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,balanced,0.2076746622721354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,balanced,0.260970671971639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,balanced,0.35997335116068524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,balanced,0.45793600877126056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,power_law_1.01,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,power_law_1.01,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,power_law_1.01,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,power_law_1.01,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,power_law_1.01,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,power_law_1.01,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,power_law_1.01,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,power_law_1.01,0.09327999949455261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,power_law_1.01,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,power_law_1.01,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.01,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.01,0.15024640560150146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.01,0.17534079551696777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.01,0.21845760345458984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.01,0.26070399284362794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.01,0.3613055944442749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.01,0.44695677757263186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.01,0.6723328113555909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.01,0.87008638381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.2,0.8524288177490235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.08869760036468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.08923519849777221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.09025920033454896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.1641535997390747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.1859071969985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.23733758926391602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.2967679977416992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.2,1.1989312171936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.41775360107421877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.5369728088378907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.7564671993255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.9828031539916993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.4256511688232423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,power_law_1.01,0.09230080246925354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,power_law_1.01,0.10929919481277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,power_law_1.01,0.15693440437316894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,power_law_1.01,0.20193281173706054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,power_law_1.01,0.2475584030151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,power_law_1.01,0.31744000911712644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,power_law_1.01,0.3287103891372681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,power_law_1.01,0.34055678844451903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,power_law_1.01,0.34953598976135253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,power_law_1.01,0.36092801094055177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,power_law_1.01,0.36985599994659424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,power_law_1.01,0.3816447973251343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,power_law_1.01,0.38455679416656496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,power_law_1.01,0.39831678867340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,power_law_1.01,0.41221117973327637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,power_law_1.01,0.4304512023925781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,power_law_1.01,0.4711872100830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.01,0.5080895900726319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.01,0.5994048118591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.01,0.6821248054504394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.01,0.8466560363769531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.01,0.9760959625244141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.01,1.303865623474121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.01,1.5885760307312011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.01,2.202572822570801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.01,2.7971263885498048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.8436992645263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,balanced,0.03230399886767069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,balanced,0.03506666670242945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,balanced,0.04935466746489207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,balanced,0.09701333443323772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,balanced,0.11890133221944173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,balanced,0.16130666931470236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,balanced,0.1946986714998881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,balanced,0.2746986746788025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,balanced,0.35210132598876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.2,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.2,0.04037120044231415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.2,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.2,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.2,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.2,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.2,0.15443840026855468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.2,0.20523519515991212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,balanced,0.03381866713364919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,balanced,0.047269334395726524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,balanced,0.08196799953778584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,balanced,0.08189333478609721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,balanced,0.08256533245245616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,balanced,0.08384533723195393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,balanced,0.08354666829109192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,balanced,0.08303466439247131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,balanced,0.08686932921409607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,balanced,0.0913759966691335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,balanced,0.09124267101287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,balanced,0.09309333562850952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,balanced,0.10518933335940044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,balanced,0.10306666294733684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,balanced,0.12082667152086894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,balanced,0.12036266922950745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,balanced,0.16541866461435953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,balanced,0.1632319986820221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,balanced,0.21981332699457803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,balanced,0.24283732970555624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,balanced,0.3447733322779338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,balanced,0.4023573398590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.2,0.3333823919296265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.2,0.4025792121887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.09935359954833985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.12679040431976318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.15571839809417726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.21059200763702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.3042112112045288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.3598207950592041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.496563196182251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,power_law_1.2,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,power_law_1.2,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,power_law_1.2,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,power_law_1.2,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,power_law_1.2,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,power_law_1.2,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,power_law_1.2,0.07885439991950989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,power_law_1.2,0.08175359964370728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,power_law_1.2,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.2,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.2,0.1248703956604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.2,0.15164799690246583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.2,0.1870911955833435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.2,0.22683520317077638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.2,0.30192639827728274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.2,0.3797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.2,0.5931392192840577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.2,0.7297599792480469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,balanced,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,balanced,0.07627733548482259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,balanced,0.07818666597207387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,balanced,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,balanced,0.07936533292134602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,balanced,0.07914133369922638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,balanced,0.0827946662902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,balanced,0.08228800197442372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,balanced,0.08397333820660909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,balanced,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,balanced,0.08639466762542725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,balanced,0.0928106705347697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,balanced,0.0937493344148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,balanced,0.09763733545939128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,balanced,0.11038399736086528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,balanced,0.12158933281898499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,balanced,0.14146133263905844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,balanced,0.16009066502253214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,balanced,0.2183893322944641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,balanced,0.25253333648045856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,balanced,0.3664640188217163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,balanced,0.45985599358876544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,balanced,0.6635146538416544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,balanced,0.860431989034017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.08837119936943054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.09938560128211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.10487040281295776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.10605440139770508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.1159168004989624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.13731839656829833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.13961600065231322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.1510975956916809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.17649919986724855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.1860479950904846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.23733758926391602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.2828416109085083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.37788798809051516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5062975883483887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.7157055854797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8192192077636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.1693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.492255973815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,power_law_1.01,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,power_law_1.01,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,power_law_1.01,0.1029312014579773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,power_law_1.01,0.10270719528198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,power_law_1.01,0.11221760511398315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,power_law_1.01,0.11412479877471923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,power_law_1.01,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,power_law_1.01,0.126636803150177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,power_law_1.01,0.13767679929733276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,power_law_1.01,0.1444159984588623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,power_law_1.01,0.14854400157928466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,power_law_1.01,0.1683776021003723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,power_law_1.01,0.18085119724273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,power_law_1.01,0.22133119106292726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,power_law_1.01,0.2657023906707764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,power_law_1.01,0.348307204246521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,power_law_1.01,0.4101503849029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,power_law_1.01,0.5711935997009278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,power_law_1.01,0.7339327812194825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,power_law_1.01,1.0371328353881837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,power_law_1.01,1.2495936393737792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.18405760526657106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.24889600276947021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.24284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.24325759410858155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.24569599628448485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.2336575984954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.2108992099761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.22533760070800782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.2440704107284546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.23096959590911864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.23840639591217042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.24904320240020753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.24924159049987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.2431936025619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.24993278980255126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.25464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.25737600326538085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.2600575923919678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.27716479301452634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.3217344045639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.35099520683288576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.39787518978118896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.4695104122161865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,power_law_1.01,0.6072127819061279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,power_law_1.01,0.7343296051025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,power_law_1.01,1.0309568405151368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,power_law_1.01,1.2324416160583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,power_law_1.2,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,power_law_1.2,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,power_law_1.2,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,power_law_1.2,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,power_law_1.2,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,power_law_1.2,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,power_law_1.2,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,power_law_1.2,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,power_law_1.2,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,power_law_1.2,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,power_law_1.2,0.05117440223693848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,power_law_1.2,0.06399999856948853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,power_law_1.2,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,power_law_1.2,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,power_law_1.2,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,power_law_1.2,0.22897920608520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,power_law_1.2,0.2881792068481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,power_law_1.2,0.41223678588867185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,power_law_1.2,0.5187647819519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,balanced,0.04911999901135763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,balanced,0.06136533121267954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,balanced,0.0639626681804657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,balanced,0.06577600042025249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,balanced,0.06765866776307423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,balanced,0.06835199892520905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,balanced,0.07155199845631917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,balanced,0.07421333094437917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,balanced,0.08892266949017842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,balanced,0.09902933239936829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,balanced,0.13477866848309836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,balanced,0.13504532972971597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,balanced,0.17469332615534464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,balanced,0.19957866271336874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,balanced,0.27685866753260296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,balanced,0.3309920032819112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,balanced,0.4733653465906779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,balanced,0.5901066859563192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,balanced,0.039408000806967415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,balanced,0.061424002051353455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,balanced,0.07235200206438701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,balanced,0.07256000240643819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,balanced,0.07392533123493195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,balanced,0.0747680018345515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,balanced,0.07530666887760162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,balanced,0.07612800101439159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,balanced,0.08146666487058003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,balanced,0.08185600241025288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,balanced,0.08504533767700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,balanced,0.08973866701126099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,balanced,0.09617066383361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,balanced,0.12109866738319397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,balanced,0.1222773293654124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,balanced,0.15624533096949259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,balanced,0.17671465873718262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,balanced,0.2420426607131958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,balanced,0.27184534072875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,balanced,0.38685333728790283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,balanced,0.47066132227579754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.07649279832839966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.10130560398101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.10954879522323609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.1332352042198181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.1475200057029724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.15850239992141724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.1482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.18410880565643312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.22288639545440675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.2815295934677124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.3304255962371826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,power_law_1.2,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,power_law_1.2,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,power_law_1.2,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,power_law_1.2,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,power_law_1.2,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,power_law_1.2,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,power_law_1.2,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,power_law_1.2,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,power_law_1.2,0.10597759485244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.4414976119995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,power_law_1.2,0.13658879995346068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,power_law_1.2,0.19256319999694824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.6297215938568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,power_law_1.2,0.2512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,0.9230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,power_law_1.2,0.3646015882492065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.3163007736206054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,power_law_1.2,0.46645121574401854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,1.8792320251464845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,2.837811279296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.01,0.05679360032081604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.01,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.01,0.05489919781684875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.01,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.01,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.01,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.01,0.08926079869270324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.01,0.10590720176696777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.01,0.1146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.01,0.14289920330047606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.01,0.15205119848251342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.01,0.20424959659576417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.01,0.21527678966522218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.01,0.27983999252319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.01,0.34961280822753904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.01,0.46975998878479003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.01,0.6081855773925782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,balanced,0.058837334314982094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,balanced,0.10161067048708598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,balanced,0.14708266655604044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,balanced,0.21945599714914957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,balanced,0.2220319906870524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,balanced,0.22806400060653687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,balanced,0.22895999749501547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,balanced,0.2300106684366862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,balanced,0.2342026631037394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,balanced,0.23675199349721274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,balanced,0.23619200785954794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,balanced,0.23733333746592203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,balanced,0.2457866668701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,balanced,0.2507359981536865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,balanced,0.2584106723467509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,balanced,0.28033600250879925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.2978453238805135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.33794132868448895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.37652798493703205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.4633386532465617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.5463893413543701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.7770026524861654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.9295573234558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,balanced,1.349071979522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,balanced,1.7468266487121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.1397312045097351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.16855679750442504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.25541119575500487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.272377610206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.28652799129486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.28714239597320557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.29161601066589354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3000063896179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.30677759647369385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.3222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.32629120349884033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.3400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.34812159538269044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.37032959461212156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.411897611618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.4765183925628662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.5732607841491699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.6484032154083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.8155967712402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.046675205230713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.3372735977172852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,power_law_1.01,1.6893760681152343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,power_law_1.01,2.238681602478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,power_law_1.01,2.7527103424072266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,power_law_1.2,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,power_law_1.2,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,power_law_1.2,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,power_law_1.2,0.10629119873046874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,power_law_1.2,0.10828160047531128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,power_law_1.2,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,power_law_1.2,0.10227199792861938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,power_law_1.2,0.1080896019935608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,power_law_1.2,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,power_law_1.2,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,power_law_1.2,0.13224960565567018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,power_law_1.2,0.13747199773788452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,power_law_1.2,0.15248639583587648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,power_law_1.2,0.17378560304641724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,power_law_1.2,0.18124799728393554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,power_law_1.2,0.22624640464782714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,power_law_1.2,0.26141440868377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,power_law_1.2,0.3428031921386719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,power_law_1.2,0.4308608055114746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,power_law_1.2,0.5867839813232422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,power_law_1.2,0.7533184051513672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,power_law_1.2,1.0760576248168945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,power_law_1.2,1.374668788909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.07924479842185975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.0794431984424591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.09510400295257568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.08751360177993775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.09107840061187744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.09820160269737244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.11082240343093872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.10568959712982177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.12352639436721802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.14339200258255005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.14454400539398193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.15111039876937865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.18536319732666015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.24028799533843995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.3175296068191528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,power_law_1.2,0.36151680946350095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,power_law_1.2,0.4859456062316895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,power_law_1.2,0.7013376235961915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.01,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.01,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.01,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.01,0.07292159795761108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.01,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.01,0.09441279768943786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.01,0.09894400238990783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.01,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.01,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.01,0.10592000484466553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.01,0.10865919589996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.01,0.11137919425964356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.01,0.12035839557647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.01,0.1202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.01,0.14566400051116943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,power_law_1.2,1.118278408050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.01,0.15999360084533693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.01,0.1909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.01,0.21928958892822265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.01,0.2755712032318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.01,0.32143359184265136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.01,0.45508480072021484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.01,0.5588160037994385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.01,0.7584191799163819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.01,0.9761343955993652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,power_law_1.2,1.313913631439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,power_law_1.2,2.0610944747924806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,balanced,0.06465599934260051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,balanced,0.0839413305123647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,balanced,0.19527467091878256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,balanced,0.33156800270080566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,balanced,0.46113598346710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,balanced,0.46541865666707355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,balanced,0.47045334180196124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,balanced,0.4709920088450114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,balanced,0.4751146634419759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,balanced,0.48314666748046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,balanced,0.48637866973876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,balanced,0.4943840106328328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,balanced,0.5049066543579102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,balanced,0.5194720029830933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,balanced,0.5293706655502319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,balanced,0.5530080000559489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,balanced,0.6149280071258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,balanced,0.6456746657689413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,balanced,0.8880053361256918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,balanced,0.8882986704508463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,balanced,1.4394133885701497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,balanced,1.4572532971700032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,balanced,2.3229494094848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,balanced,2.439466635386149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,balanced,3.722581227620443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,balanced,4.664981206258138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.09227520227432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.12889599800109863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.16284799575805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.17809280157089233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.2118016004562378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.21783039569854737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.2272576093673706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.2347327947616577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.24849920272827147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.25284481048583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,power_law_1.2,2.838630485534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2658112049102783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.2753472089767456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.28343679904937746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.30103681087493894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.39149439334869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.48122239112854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.5774144172668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7443071842193604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.9300288200378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.263475227355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.4846400260925292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.1227903366088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.7416383743286135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,power_law_1.01,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,power_law_1.01,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,power_law_1.01,0.07562239766120911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,power_law_1.01,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,power_law_1.01,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,power_law_1.01,0.09240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,power_law_1.01,0.1046336054801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,power_law_1.01,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,power_law_1.01,0.14264960289001466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,power_law_1.01,0.1733247995376587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,power_law_1.01,0.2079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,power_law_1.01,0.2704960107803345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,power_law_1.01,0.32928640842437745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,power_law_1.01,0.48156161308288575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,power_law_1.01,0.5904895782470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,power_law_1.01,0.850438404083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,power_law_1.01,1.1459136009216309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,balanced,0.05885333319505056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,balanced,0.05871466795603434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,balanced,0.06252266466617584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,balanced,0.06357333560784657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,balanced,0.06497066716353099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,balanced,0.06984533369541168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,balanced,0.07061333457628886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,balanced,0.07334400216738383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,balanced,0.09130133191744487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,balanced,0.11953066786130269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,balanced,0.1360373298327128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,balanced,0.1822026570638021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,balanced,0.16375466187795004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,balanced,0.21085333824157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,balanced,0.2453493277231852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,balanced,0.33318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,balanced,0.3951146602630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.10331519842147827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.11440000534057618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.15893759727478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.18511999845504762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.20404479503631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.23781120777130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.2482815980911255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.25846400260925295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.2673791885375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.27653119564056394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.2789376020431519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.27818880081176756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.2836287975311279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.3028480052947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.3166975975036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.3376895904541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.37441918849945066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.42807679176330565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.4756864070892334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.6137792110443115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.6925824165344239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,0.920582389831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.163967990875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,1.6181312561035157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.000774383544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,balanced,0.04828799764315287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,balanced,0.05726933479309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,balanced,0.08049599826335907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,balanced,0.11250133315722148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,balanced,0.14245866735776266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,balanced,0.14378666877746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,balanced,0.14410133163134256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,balanced,0.1418293317159017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,balanced,0.1455946664015452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,balanced,0.14566933115323386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,balanced,0.15038933356602988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,balanced,0.15938666462898254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,balanced,0.16060266892115274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,balanced,0.17283199230829874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,balanced,0.1918720006942749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,balanced,0.21242133776346842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,balanced,0.24673600991566977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,balanced,0.2773759961128235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,balanced,0.3428106705347697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,balanced,0.4306666851043701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,balanced,0.5624320109685262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,balanced,0.7262933254241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,balanced,1.012768030166626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,balanced,1.2849173545837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,balanced,0.04610666632652283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,balanced,0.06890133519967397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,balanced,0.07054933408896129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,balanced,0.07259733478228252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,balanced,0.07226666808128357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,balanced,0.07493866483370464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,balanced,0.07598400115966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,balanced,0.0774239997069041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,balanced,0.09079466263453166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,balanced,0.0934879978497823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,balanced,0.09155199925104777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,balanced,0.10081600149472554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,balanced,0.10537599523862202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,balanced,0.12592533230781555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,balanced,0.1365386644999186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,balanced,0.1681013305981954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,balanced,0.1888373295466105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,balanced,0.24267733097076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,balanced,0.2919519941012065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,balanced,0.4017866849899292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,balanced,0.53056534131368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,power_law_1.01,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,power_law_1.01,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,power_law_1.01,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,power_law_1.01,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,power_law_1.01,0.058931201696395874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,power_law_1.01,0.07815679907798767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,power_law_1.01,0.09656320214271545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,power_law_1.01,0.10668799877166749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,power_law_1.01,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,power_law_1.01,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,power_law_1.01,0.22198400497436524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,power_law_1.01,0.2690752029418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,power_law_1.01,0.36957440376281736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,power_law_1.01,0.46389122009277345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13098880052566528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.13670400381088257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.1423807978630066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.14495999813079835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.14547840356826783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.15426559448242189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.1612671971321106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.16364799737930297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.1724544048309326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.18373759984970092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.19405440092086793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.20490241050720215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.24851200580596924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.2877311944961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.3695231914520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.45476479530334474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.599129581451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7179647922515869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,power_law_1.2,0.9744895935058594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.3102784156799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,balanced,0.047279998660087585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,balanced,0.0658133327960968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,balanced,0.08731200297673543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,balanced,0.08734933535257976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,balanced,0.09186133742332458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,balanced,0.09195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,balanced,0.09962133566538493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,balanced,0.10244266192118327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,balanced,0.11188800136248271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,balanced,0.12325333555539449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,balanced,0.13505066434542337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,balanced,0.15875200430552164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,balanced,0.18013866742451987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,balanced,0.22616000970204672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,balanced,0.28812267382939655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,balanced,0.3864426612854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,balanced,0.4975626468658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,balanced,0.7125546932220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.7974592208862306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,balanced,0.9030506610870361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.285196876525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,power_law_1.01,0.061363202333450315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,power_law_1.01,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,power_law_1.01,0.08214399814605713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,power_law_1.01,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,power_law_1.01,0.10219520330429077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,power_law_1.01,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,power_law_1.01,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,power_law_1.01,0.11395839452743531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,power_law_1.01,0.11634559631347656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,power_law_1.01,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,power_law_1.01,0.12106239795684814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,power_law_1.01,0.126803195476532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,power_law_1.01,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,power_law_1.01,0.1399616003036499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,power_law_1.01,0.15599360466003417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.01,0.17175040245056153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.01,0.2063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.01,0.24641919136047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.01,0.31911680698394773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.01,0.39095680713653563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.01,0.5524608135223389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.01,0.6876351833343506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.01,1.0357248306274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.01,1.259488010406494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,balanced,0.05444266895453135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,balanced,0.05532266696294149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,balanced,0.06934933364391327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,balanced,0.07978666822115581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,balanced,0.08221866687138875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,balanced,0.08100800216197968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,balanced,0.0812853326400121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,balanced,0.08241066833337148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,balanced,0.08347200353940327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,balanced,0.08423999945322673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,balanced,0.0897813340028127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,balanced,0.08987200260162354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,balanced,0.09319466352462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,balanced,0.10917333761850993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,balanced,0.10827733079592387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,balanced,0.13402133186658224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,balanced,0.13981333374977112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,balanced,0.17909866571426392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,balanced,0.17437867323557535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,balanced,0.21916800737380981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,balanced,0.23457600673039755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,balanced,0.3130826751391093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,balanced,0.36236266295115155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.2,0.02768000066280365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.2,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.2,0.028960001468658448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.2,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.2,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.2,0.029235199093818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.2,0.029791998863220214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.2,0.030195200443267824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.2,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.2,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.2,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.2,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.2,0.1619328022003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.2,0.2182528018951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.2,0.2738624095916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,power_law_1.2,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,power_law_1.2,0.08260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,power_law_1.2,0.09278079867362976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,power_law_1.2,0.10369919538497925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,power_law_1.2,0.1026304006576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,power_law_1.2,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,power_law_1.2,0.10862079858779908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,power_law_1.2,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,power_law_1.2,0.11037440299987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,power_law_1.2,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,power_law_1.2,0.12490240335464478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,power_law_1.2,0.1288256049156189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,power_law_1.2,0.1392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,power_law_1.2,0.15404800176620484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,power_law_1.2,0.16451200246810913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,power_law_1.2,0.20076799392700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,power_law_1.2,0.21753599643707275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,power_law_1.2,0.2817471981048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,power_law_1.2,0.3277184009552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,power_law_1.2,0.45549440383911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,power_law_1.2,0.5567488193511962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,power_law_1.2,0.8078720092773437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,power_law_1.2,0.9671872138977051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,power_law_1.01,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,power_law_1.01,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,power_law_1.01,0.03635840117931366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,power_law_1.01,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,power_law_1.01,0.0423552006483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,power_law_1.01,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,power_law_1.01,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,power_law_1.01,0.08019199967384338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.01,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.01,0.12666239738464355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.01,0.15293439626693725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.01,0.2112191915512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.01,0.2835968017578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.01,0.4292992115020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.01,0.5585343837738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.01,0.9253503799438476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.06318719983100891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.01,1.1837759971618653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.08517760038375854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.09315199851989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.12791039943695068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.1250175952911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.14926079511642457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.1611583948135376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.21235198974609376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.28682880401611327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,power_law_1.2,0.3692352056503296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,power_law_1.2,0.4644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,power_law_1.2,0.7828671932220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,balanced,0.02938666691382726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,balanced,0.029552000264326733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,balanced,0.03499733408292135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,balanced,0.034448000291983284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,balanced,0.035989334185918175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,balanced,0.03806400050719579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,balanced,0.041482667128245033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,balanced,0.06164266665776571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,balanced,0.07607466479142506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,balanced,0.09103999535242717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,balanced,0.11145599683125813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,balanced,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,power_law_1.2,1.159001636505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,power_law_1.01,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,power_law_1.01,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,power_law_1.01,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,power_law_1.01,0.05736320018768311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,power_law_1.01,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,power_law_1.01,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,power_law_1.01,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,power_law_1.01,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,power_law_1.01,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,power_law_1.01,0.10525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,power_law_1.01,0.11978880167007447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,power_law_1.01,0.15012480020523072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,power_law_1.01,0.17149440050125123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,power_law_1.01,0.22637441158294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,power_law_1.01,0.2931904077529907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,power_law_1.01,0.40684800148010253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,power_law_1.01,0.4755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,power_law_1.01,0.6650368213653565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,power_law_1.01,0.9302847862243653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,power_law_1.01,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,power_law_1.01,0.09648000001907349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,power_law_1.01,0.1004863977432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,power_law_1.01,0.10529279708862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,power_law_1.01,0.10657919645309448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,power_law_1.01,0.10177919864654542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,power_law_1.01,0.11006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,power_law_1.01,0.11021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,power_law_1.01,0.11152640581130982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,power_law_1.01,0.11177599430084229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,power_law_1.01,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,power_law_1.01,0.11733759641647339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,power_law_1.01,0.1498047947883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.01,0.1704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.01,0.21088640689849852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.01,0.24412159919738768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.01,0.33214080333709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.01,0.3669248104095459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.01,0.5606207847595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.01,0.6536575794219971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.01,0.9925567626953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.01,1.2577664375305175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,power_law_1.01,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,power_law_1.01,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,power_law_1.01,0.06747519969940186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,power_law_1.01,0.1359231948852539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,power_law_1.01,0.1955199956893921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,power_law_1.01,0.22907519340515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,power_law_1.01,0.3385215997695923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,power_law_1.01,0.4268608093261719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,power_law_1.01,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,power_law_1.01,0.07824640274047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,power_law_1.01,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,power_law_1.01,0.08318719863891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,power_law_1.01,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,power_law_1.01,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,power_law_1.01,0.09863680005073547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,power_law_1.01,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,power_law_1.01,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,power_law_1.01,0.12375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,power_law_1.01,0.13300479650497438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,power_law_1.01,0.16472959518432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,power_law_1.01,0.19073280096054077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,power_law_1.01,0.24194560050964356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,power_law_1.01,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,power_law_1.01,0.40853118896484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,power_law_1.01,0.53471360206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,power_law_1.01,0.729804801940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,power_law_1.01,1.0089471817016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,balanced,0.05948266883691152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,balanced,0.06243200103441874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,balanced,0.0633653352657954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,balanced,0.06373866895834605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,balanced,0.06473599870999654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,balanced,0.06966933111349742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,balanced,0.07665066421031952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,balanced,0.07966400186220805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,balanced,0.08353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,balanced,0.09448533256848653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,balanced,0.10608533024787903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,balanced,0.12480533123016357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,balanced,0.1402239998181661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,balanced,0.1869973341623942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,balanced,0.21795199314753214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,balanced,0.3235893249511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,balanced,0.4102880160013835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,balanced,0.6083466609319051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,balanced,0.8124213218688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,power_law_1.01,0.10387200117111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,power_law_1.01,0.13481600284576417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,power_law_1.01,0.144486403465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,power_law_1.01,0.18148479461669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,power_law_1.01,0.21228160858154296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,power_law_1.01,0.2586047887802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,power_law_1.01,0.289683198928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,power_law_1.01,0.28853759765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,power_law_1.01,0.2841023921966553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,power_law_1.01,0.2859711885452271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,power_law_1.01,0.29656319618225097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,power_law_1.01,0.30986239910125735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,power_law_1.01,0.3042304039001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,power_law_1.01,0.3172096014022827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,power_law_1.01,0.3302527904510498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,power_law_1.01,0.34792320728302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,power_law_1.01,0.3521087884902954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,power_law_1.01,0.3944511890411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,power_law_1.01,0.41792640686035154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,power_law_1.01,0.4940927982330322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,power_law_1.01,0.5569983959197998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,power_law_1.01,0.6678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,power_law_1.01,0.725548791885376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,power_law_1.01,0.9129792213439941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,power_law_1.01,1.111628818511963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,power_law_1.01,1.5385919570922852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,power_law_1.01,1.9913536071777345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,8,power_law_1.01,0.21551361083984374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,8,power_law_1.01,0.3307264089584351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,8,power_law_1.01,0.29887359142303466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,8,power_law_1.01,0.34281599521636963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,8,power_law_1.01,0.3908224105834961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,8,power_law_1.01,0.3804480075836182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,8,power_law_1.01,0.41673598289489744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,8,power_law_1.01,0.3939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,8,power_law_1.01,0.3868864059448242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,8,power_law_1.01,0.4037631988525391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,8,power_law_1.01,0.4018752098083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,8,power_law_1.01,0.4076672077178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,8,power_law_1.01,0.4181375980377197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,8,power_law_1.01,0.4353856086730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,8,power_law_1.01,0.429856014251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,8,power_law_1.01,0.4387968063354492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,8,power_law_1.01,0.45046401023864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,8,power_law_1.01,0.4824384212493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.01,0.5177279949188233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.01,0.5878015995025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.01,0.6624896049499511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.01,0.852723217010498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.01,1.0194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.01,1.341478443145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.01,1.7098175048828126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.01,2.4739263534545897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.01,3.0635135650634764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.1098431944847107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.10792959928512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.15393919944763185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.19119999408721924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.18977919816970826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.25471999645233157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.25311999320983886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.25160961151123046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.26363520622253417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.2708928108215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.26346240043640134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.2887615919113159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.28088960647583006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.30001280307769773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.30353920459747313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.31918718814849856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.336627197265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.3708928108215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.44824957847595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4851776123046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.5618239879608155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.6579135894775391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.8974464416503907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.024550437927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.5644864082336425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,1.8824127197265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,power_law_1.01,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,power_law_1.01,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,power_law_1.01,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,power_law_1.01,0.09219840168952942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,power_law_1.01,0.09468160271644592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,power_law_1.01,0.09159039855003356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,power_law_1.01,0.09700480103492737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,power_law_1.01,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,power_law_1.01,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,power_law_1.01,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,power_law_1.01,0.10120320320129395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,power_law_1.01,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,power_law_1.01,0.10570240020751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,power_law_1.01,0.10794240236282349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,power_law_1.01,0.11812479496002197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.01,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.01,0.14903680086135865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.01,0.1677183985710144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.01,0.20639359951019287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.01,0.24812159538269044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.01,0.33971199989318845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.01,0.4283455848693848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.01,0.6415232181549072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.01,0.8782400131225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.029996800422668456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.03805440068244934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.15357439517974852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.19617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.2291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,balanced,0.03814399987459183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,balanced,0.033914667864640556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,balanced,0.03383466601371765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,balanced,0.03851199895143509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,balanced,0.0448586642742157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,balanced,0.08146133522192638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,balanced,0.10331199566523235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,balanced,0.12139200170834859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.3310528039932251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,power_law_1.2,0.04177280068397522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,power_law_1.2,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.2,0.07684479951858521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.2,0.09189119935035706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.2,0.12309759855270386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.2,0.13600000143051147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.2,0.19969279766082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.2,0.22632319927215577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,power_law_1.2,0.40038399696350097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.2,0.3328768014907837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.2,0.5030848026275635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,power_law_1.2,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,power_law_1.2,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,power_law_1.2,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,power_law_1.2,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,power_law_1.2,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,power_law_1.2,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,power_law_1.2,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,power_law_1.2,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,power_law_1.2,0.10554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,power_law_1.2,0.14254080057144164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,power_law_1.2,0.17889920473098755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,power_law_1.2,0.24511361122131348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,power_law_1.2,0.30471038818359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,power_law_1.2,0.41532158851623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,power_law_1.2,0.5603392124176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,power_law_1.2,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,power_law_1.2,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,power_law_1.2,0.07196159958839417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,power_law_1.2,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,power_law_1.2,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,power_law_1.2,0.07844480276107788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,power_law_1.2,0.08629760146141052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,power_law_1.2,0.09824640154838563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,power_law_1.2,0.10733439922332763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.2,0.12040959596633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.2,0.14828159809112548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.2,0.1716928005218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.2,0.24259839057922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.2,0.3186048030853271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.2,0.43674240112304685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.2,0.6599103927612304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.2,1.0394368171691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.2,1.4598272323608399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,balanced,0.07417599856853485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,balanced,0.08878933389981587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,balanced,0.08734400073687236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,balanced,0.0872213343779246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,balanced,0.0888853371143341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,balanced,0.09107733766237895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,balanced,0.091648002465566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,balanced,0.097680002450943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,balanced,0.09691733121871948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,balanced,0.10157333811124165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,balanced,0.1072160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,balanced,0.11133866508801778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,balanced,0.1236853301525116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,balanced,0.1336373289426168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,balanced,0.15607999761899313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,balanced,0.17892267306645712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,balanced,0.2187946637471517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,balanced,0.2683733304341634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,balanced,0.35018134117126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,balanced,0.44366931915283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,balanced,0.027802666028340656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,balanced,0.03183466692765554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,balanced,0.03230933348337809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,balanced,0.04409066836039225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,balanced,0.04242133100827535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,balanced,0.04966400067011515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,balanced,0.06517866750558217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,balanced,0.07877866427103679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,balanced,0.10430399576822917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,balanced,0.11122133334477742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,balanced,0.15180266896883646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,balanced,0.16790932416915894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,balanced,0.234442671140035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,balanced,0.30929599205652875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.08784000277519226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.0944320023059845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.09388160109519958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.11847679615020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.15159679651260377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.167193603515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.22515199184417725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.2603391885757446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.365120005607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,power_law_1.2,0.45691518783569335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,power_law_1.2,0.6845183849334717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,power_law_1.2,0.8715456008911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,4,power_law_1.2,0.13864959478378297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,4,power_law_1.2,0.15930880308151246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,4,power_law_1.2,0.18893439769744874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,4,power_law_1.2,0.20361599922180176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,4,power_law_1.2,0.23845760822296141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,4,power_law_1.2,0.3266304016113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,4,power_law_1.2,0.33792641162872317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,4,power_law_1.2,0.3195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,4,power_law_1.2,0.3532095909118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,4,power_law_1.2,0.35586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,4,power_law_1.2,0.3697792053222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,4,power_law_1.2,0.37509760856628416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,4,power_law_1.2,0.37356159687042234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,4,power_law_1.2,0.3813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,4,power_law_1.2,0.40561280250549314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,4,power_law_1.2,0.39733119010925294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,4,power_law_1.2,0.3998656034469604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,4,power_law_1.2,0.4440192222595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.2,0.47660160064697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.2,0.543507194519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.2,0.6303423881530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.2,0.7837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.2,0.9446208000183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.2,1.3537983894348145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.2,1.840358352661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.2,2.4942975997924806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,balanced,0.027642667293548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,balanced,0.027893332143624622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,balanced,0.03121600051720937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,balanced,0.034917332231998444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,balanced,0.042133331298828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,balanced,0.05108266572157542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,balanced,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,balanced,0.09139200051625569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,balanced,0.1134986678759257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,balanced,0.141866664091746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,balanced,0.1733120083808899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.2,3.4172672271728515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.08793600201606751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.09731839895248413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.11845760345458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.14071680307388307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.14143359661102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.15237120389938355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.14761600494384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.1564288020133972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.15992319583892822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.1629439949989319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.1689087986946106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.1763327956199646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.18061439990997313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.2160576105117798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.24520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.28538239002227783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.34876160621643065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.450054407119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.5537216186523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.7401472091674804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,0.9902848243713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,1.3997440338134766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,1.8442432403564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,power_law_1.2,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,power_law_1.2,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,power_law_1.2,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,power_law_1.2,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,power_law_1.2,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,power_law_1.2,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,power_law_1.2,0.09265279769897461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,power_law_1.2,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,power_law_1.2,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,power_law_1.2,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,power_law_1.2,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,power_law_1.2,0.12060799598693847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,power_law_1.2,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,power_law_1.2,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.2,0.15360000133514404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.2,0.19711359739303588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.2,0.22446079254150392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.2,0.3089087963104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.2,0.3605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.2,0.5436031818389893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.2,0.7423295974731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.2,1.0728832244873048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.2,1.3164031982421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,power_law_1.01,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,power_law_1.01,0.07802879810333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,power_law_1.01,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,power_law_1.01,0.0730239987373352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,power_law_1.01,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,power_law_1.01,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,power_law_1.01,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,power_law_1.01,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,power_law_1.01,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,power_law_1.01,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,power_law_1.01,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,power_law_1.01,0.1288256049156189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,power_law_1.01,0.14327679872512816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,power_law_1.01,0.18243839740753173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,power_law_1.01,0.2101952075958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,power_law_1.01,0.28255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,power_law_1.01,0.3469887971878052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,power_law_1.01,0.48836479187011717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,power_law_1.01,0.6415679931640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,balanced,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,balanced,0.05667733152707418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,balanced,0.05967999994754791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,balanced,0.08388800422350566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,balanced,0.08160000046094258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,balanced,0.08292266726493835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,balanced,0.08061866462230682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,balanced,0.08283199866612752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,balanced,0.07824533184369405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,balanced,0.08349333206812541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,balanced,0.08499200145403545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,balanced,0.08691733082135518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,balanced,0.09159466624259949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,balanced,0.09506666660308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,balanced,0.10009599725405376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.1032426655292511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.11646399895350139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.12905066212018332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.15422399838765463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.1778293251991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.23460266987482706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.27774399518966675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.3834933439890544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.48499735196431476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,balanced,0.06206400195757548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,balanced,0.06411733229955037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,balanced,0.07398400207360585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,balanced,0.13223466277122498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,balanced,0.13236266374588013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,balanced,0.1317759950955709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,balanced,0.13111467162768045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,balanced,0.13379733761151633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,balanced,0.13542933265368143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,balanced,0.13807466626167297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,balanced,0.13955199718475342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,balanced,0.142277330160141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,balanced,0.1477226714293162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,balanced,0.1490613321463267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,balanced,0.15566399693489075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,balanced,0.17483200629552206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,balanced,0.18826133012771606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,balanced,0.2193280061086019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,balanced,0.24974934260050455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,balanced,0.3149706721305847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,balanced,0.38072001934051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,balanced,0.544485330581665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,balanced,0.6699999968210856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,balanced,0.9776906967163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,balanced,1.2642186482747395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.01,0.06839680075645446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.01,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.01,0.08492159843444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.01,0.08895360231399536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.01,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.01,0.10390399694442749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.01,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.01,0.10634880065917969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.01,0.10369919538497925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.01,0.10412160158157349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.01,0.11023999452590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.01,0.11458560228347778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.01,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.01,0.12684160470962524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.01,0.129203200340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.01,0.13481600284576417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.01,0.15569280385971068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.01,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.01,0.21628160476684571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.01,0.2614464044570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.01,0.33656959533691405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.01,0.3634943962097168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.01,0.4841343879699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.01,0.5768767833709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.01,0.8702655792236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.01,1.093183994293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.05968639850616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.06046079993247986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.0689408004283905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.08104959726333619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.10179840326309204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.12440320253372192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.1673535943031311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.23582079410552978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.27708799839019777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.39288959503173826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.5595647811889648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.820678424835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,balanced,0.0554666668176651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,balanced,0.055717334151268005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,balanced,0.055861334005991616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,balanced,0.05787200232346853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,balanced,0.05917333563168844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,balanced,0.059088001648585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,balanced,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,balanced,0.06374933322270711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,balanced,0.06494933366775513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,balanced,0.0727893312772115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,balanced,0.07508266468842824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,balanced,0.08181333541870117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,balanced,0.09002133210500081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,balanced,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,balanced,0.12135466933250427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,balanced,0.1463093360265096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,balanced,0.19058666626612344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,balanced,0.23454399903615317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,balanced,0.3203786611557007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,balanced,0.39475735028584796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,balanced,0.5683146715164185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,balanced,0.7281493345896403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,0.9554240226745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,balanced,0.06786666810512543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,balanced,0.07898133496443431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,balanced,0.10625599821408589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,balanced,0.22299732764561972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,balanced,0.2635093331336975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,balanced,0.26942400137583417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,balanced,0.27403199672698975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,balanced,0.2780906756718953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,balanced,0.2768746614456177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,balanced,0.28355199098587036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,balanced,0.28803733984629315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,balanced,0.28654932975769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,balanced,0.29811733961105347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,balanced,0.30563199520111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,balanced,0.31285866101582843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,balanced,0.32850666840871173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,balanced,0.3712906837463379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.39908266067504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.4906719923019409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.5466773509979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.7708586851755778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.8943626880645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,balanced,1.2778080304463704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,balanced,1.622144063313802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,balanced,2.4592107137044272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,balanced,3.033664067586263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,balanced,0.033146666983763375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,balanced,0.0742986649274826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,balanced,0.1251520017782847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,balanced,0.1204266647497813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,balanced,0.12032533685366313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,balanced,0.11994666854540507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,balanced,0.12007466952006023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,balanced,0.11901866396268208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,balanced,0.12277332941691081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,balanced,0.1223413348197937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,balanced,0.12243200341860454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,balanced,0.12372266252835591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,balanced,0.12748266259829202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,balanced,0.12960533301035562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,balanced,0.13241599996884665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,balanced,0.13209600249926248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,balanced,0.13924266894658408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,balanced,0.1564533313115438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,balanced,0.15705066919326782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,balanced,0.18566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,balanced,0.1935466726620992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,balanced,0.2616960008939107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,balanced,0.2888960043589274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,balanced,0.37172265847524005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,balanced,0.46989866097768146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.059520000219345094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.12077440023422241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.14891519546508789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.21252479553222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.08925439715385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.11371519565582275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.1089087963104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.14944000244140626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.158406400680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.1654144048690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.15567359924316407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.16471680402755737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.15839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.15981440544128417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.16627199649810792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.1729856014251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.1868607997894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.18350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.19661439657211305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.2179327964782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.23400321006774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.2696768045425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.2997888088226318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.35856640338897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.43276801109313967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.567468786239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.7138432025909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,0.9970303535461426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.2518207550048828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,balanced,0.09041066964467366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,balanced,0.13648000359535217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,balanced,0.14917866388956705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,balanced,0.1546239952246348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,balanced,0.15552000204722086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,balanced,0.160181333621343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,balanced,0.1614346702893575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,balanced,0.16645333170890808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,balanced,0.17061867316563925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,balanced,0.1740106741587321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,balanced,0.18449600537618002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,balanced,0.19163199265797934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,balanced,0.19984533389409384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,balanced,0.21284266312917074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,balanced,0.254202663898468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,balanced,0.27801066637039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,balanced,0.3940853277842204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,balanced,0.4254879951477051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,balanced,0.631989320119222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,balanced,0.7267733414967855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.2620863914489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,balanced,1.124069372812907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,balanced,1.3403573036193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,balanced,2.0635093053181968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,balanced,2.5724426905314126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,power_law_1.01,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,power_law_1.01,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,power_law_1.01,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,power_law_1.01,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,power_law_1.01,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,power_law_1.01,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,power_law_1.01,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,power_law_1.01,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,power_law_1.01,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,power_law_1.01,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,power_law_1.01,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,power_law_1.01,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,power_law_1.01,0.1099776029586792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,power_law_1.01,0.14186880588531495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,power_law_1.01,0.17142399549484252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,power_law_1.01,0.23585278987884523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,power_law_1.01,0.29649920463562013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,power_law_1.01,0.3987776041030884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,power_law_1.01,0.5347775936126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,power_law_1.01,0.06332160234451294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,power_law_1.01,0.0837119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,power_law_1.01,0.13435519933700563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,power_law_1.01,0.17966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,power_law_1.01,0.1881152033805847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,power_law_1.01,0.18584320545196534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,power_law_1.01,0.19503999948501588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,power_law_1.01,0.19835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,power_law_1.01,0.20383360385894775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,power_law_1.01,0.2038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,power_law_1.01,0.2102976083755493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,power_law_1.01,0.22012801170349122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,power_law_1.01,0.23276159763336182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,power_law_1.01,0.2364351987838745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,power_law_1.01,0.24030721187591553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,power_law_1.01,0.2659327983856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,power_law_1.01,0.29000320434570315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,power_law_1.01,0.3570048093795776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,power_law_1.01,0.37389440536499025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,power_law_1.01,0.4839104175567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,power_law_1.01,0.5503488063812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,power_law_1.01,0.7584832191467286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,power_law_1.01,0.9525823593139648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,power_law_1.01,1.2774527549743653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,power_law_1.01,1.6551168441772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,balanced,0.05809600154558817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,balanced,0.06042666733264923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,balanced,0.08566400408744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,balanced,0.103685329357783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,balanced,0.10504532853762309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,balanced,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,balanced,0.10451199611028035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,balanced,0.10802132884661357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,balanced,0.10762133200963338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,balanced,0.10922666390736897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,balanced,0.1114453375339508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,balanced,0.11170132954915364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,balanced,0.12008532881736755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,balanced,0.1225386659304301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,balanced,0.12532800436019897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,balanced,0.13222400347391763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,balanced,0.13894933462142944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,balanced,0.16061866283416748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,balanced,0.17558934291203818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,balanced,0.21919999519983926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,balanced,0.25862399737040204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,balanced,0.3802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,balanced,0.4668320020039876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,balanced,0.6572693188985189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,balanced,0.8587413628896078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,power_law_1.01,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,power_law_1.01,0.06940799951553345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,power_law_1.01,0.0856000006198883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,power_law_1.01,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,power_law_1.01,0.10616320371627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,power_law_1.01,0.11217279434204101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,power_law_1.01,0.12225919961929321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,power_law_1.01,0.1272447943687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,power_law_1.01,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,power_law_1.01,0.17182719707489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.01,0.1982143998146057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.01,0.25226240158081054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.01,0.3304960012435913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.01,0.4108543872833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.01,0.5940159797668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.01,0.85731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.01,0.9462016105651856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.01,1.5796095848083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.01,1.744870376586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,power_law_1.01,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,power_law_1.01,0.09091200232505799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,power_law_1.01,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,power_law_1.01,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,power_law_1.01,0.15025919675827026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,power_law_1.01,0.19059840440750123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,power_law_1.01,0.19629440307617188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,power_law_1.01,0.20112640857696534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,power_law_1.01,0.20472960472106932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,power_law_1.01,0.20894720554351806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,power_law_1.01,0.21804161071777345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,power_law_1.01,0.22087678909301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,power_law_1.01,0.22600319385528564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,power_law_1.01,0.23205759525299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,power_law_1.01,0.2393728017807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,power_law_1.01,0.24797439575195312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,power_law_1.01,0.27042560577392577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.01,0.2926399946212769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.01,0.3460671901702881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.01,0.3900608062744141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.01,0.4907199859619141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.01,0.559929609298706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.01,0.7315455913543701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.01,0.921996784210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.01,1.2812031745910644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.01,1.6772800445556642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,balanced,0.029088000456492107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,balanced,0.027471999327341717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,balanced,0.027488000690937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,balanced,0.037274666130542755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,balanced,0.03787733366092046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,balanced,0.06407466530799866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,balanced,0.08152533570925395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,balanced,0.10588266452153523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,balanced,0.13100799918174744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,balanced,0.15494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,balanced,0.2169546683629354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,balanced,0.26316799720128375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,balanced,0.0687360018491745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,balanced,0.09251733620961507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,balanced,0.18630399306615195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,balanced,0.3120159904162089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,balanced,0.37959468364715576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,balanced,0.3785173495610555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,balanced,0.3858826557795207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,balanced,0.3880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,balanced,0.3892480134963989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,balanced,0.3890026807785034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,balanced,0.3933493296305339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,balanced,0.3970400094985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,balanced,0.40302932262420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,balanced,0.4086560010910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,balanced,0.41529067357381183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,balanced,0.4243679841359456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,balanced,0.4484533468882243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.46886932849884033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.5897333224614462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.5624213218688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.7755626837412516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.7681439717610677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.163157304128011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.3047680060068767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,balanced,1.8611680666605632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,balanced,2.369056065877279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.08709120154380798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.13745919466018677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.272870397567749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.3370752096176147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.4419519901275635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.45698561668396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.4782656192779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4844031810760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.5053952217102051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5105023860931397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5358143806457519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5441664218902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5665664196014404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5781184196472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.6235328197479248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6914368152618409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7564032077789307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8892031669616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9123007774353027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.13306884765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.3602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,1.7625728607177735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.165798377990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,2.964025688171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,3.772684860229492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,power_law_1.2,0.11038719415664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,power_law_1.2,0.13459839820861816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,power_law_1.2,0.19050240516662598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,power_law_1.2,0.25207679271697997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,power_law_1.2,0.2969791889190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,power_law_1.2,0.4540095806121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,power_law_1.2,0.4294591903686523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,power_law_1.2,0.42989439964294435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,power_law_1.2,0.4620031833648682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,power_law_1.2,0.4395711898803711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,power_law_1.2,0.4656191825866699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,power_law_1.2,0.49370880126953126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,power_law_1.2,0.5119423866271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,power_law_1.2,0.5266176223754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,power_law_1.2,0.5197887897491456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,power_law_1.2,0.5607615947723389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,power_law_1.2,0.6254848003387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,power_law_1.2,0.631712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,power_law_1.2,0.6932735919952393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,power_law_1.2,0.7337152004241944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,power_law_1.2,0.879155158996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,power_law_1.2,0.89683837890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,power_law_1.2,1.1039104461669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,power_law_1.2,1.2962559700012206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,power_law_1.2,1.6645696640014649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,power_law_1.2,2.121811294555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,4,balanced,0.09612799684206645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,4,balanced,0.10735999544461568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,4,balanced,0.11602666974067688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,4,balanced,0.1532426675160726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,4,balanced,0.2075093388557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,4,balanced,0.3192053238550822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,4,balanced,0.4352266788482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,4,balanced,0.4373386700948079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,4,balanced,0.43513067563374835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,4,balanced,0.43933331966400146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,4,balanced,0.43886931737263996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,4,balanced,0.44234665234883624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,4,balanced,0.4442453384399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,4,balanced,0.44760533173878986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,4,balanced,0.4552160104115804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,4,balanced,0.4586026668548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,4,balanced,0.4667359987894694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,4,balanced,0.48024535179138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,4,balanced,0.49604801336924237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,4,balanced,0.5321226517359415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,4,balanced,0.5729066530863444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,4,balanced,0.6494079828262329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,4,balanced,0.7237546443939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,4,balanced,0.8715519905090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,4,balanced,1.1401387055714924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,4,balanced,1.4384640057881672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,4,balanced,1.896506627400716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.01,0.08637440204620361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.01,0.12213120460510254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.01,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.01,0.14994560480117797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.01,0.17694720029830932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.01,0.2130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.01,0.21856000423431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.01,0.22812800407409667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.01,0.2262655973434448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.01,0.2329535961151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.01,0.231443190574646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.01,0.2485055923461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.01,0.24106879234313966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.01,0.2511679887771606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.01,0.26387200355529783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.01,0.2670464038848877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.01,0.2879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.01,0.32988801002502444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.01,0.34306559562683103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.01,0.4037504196166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.01,0.46201601028442385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.01,0.5851200103759766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.01,0.6075776100158692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.01,0.8001407623291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.01,1.0565631866455079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.01,1.435315227508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.01,1.7656896591186524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.2,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.2,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.2,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.2,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.2,0.08763520121574402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.2,0.08912000060081482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.2,0.09274240136146546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.2,0.10026240348815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.2,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.2,0.10417920351028442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.2,0.12332799434661865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.2,0.13562239408493043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.2,0.17127039432525634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.2,0.20641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.2,0.27382400035858157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.2,0.28433918952941895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.2,0.3906752109527588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.2,0.4791103839874268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.2,0.6959487915039062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.2,0.908083152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,power_law_1.2,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,power_law_1.2,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,power_law_1.2,0.07234560251235962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,power_law_1.2,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,power_law_1.2,0.10676480531692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,power_law_1.2,0.09896960258483886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,power_law_1.2,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,power_law_1.2,0.10817279815673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,power_law_1.2,0.10907520055770874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,power_law_1.2,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,power_law_1.2,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,power_law_1.2,0.12273279428482056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,power_law_1.2,0.1286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,power_law_1.2,0.14442239999771117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,power_law_1.2,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,power_law_1.2,0.16862720251083374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,power_law_1.2,0.2030656099319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,power_law_1.2,0.2564671993255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,power_law_1.2,0.2857408046722412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,power_law_1.2,0.36788480281829833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,power_law_1.2,0.4570496082305908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,power_law_1.2,0.6294015884399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,power_law_1.2,0.8336447715759278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.06136959791183472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.09947519898414611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.14816000461578369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.17918720245361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.26380798816680906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.293939208984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.4805439949035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.6578303813934326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,balanced,0.045514668027559914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,balanced,0.052229334910710655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,balanced,0.07025066514809926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,balanced,0.0828906645377477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,balanced,0.09215466181437175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,balanced,0.12144000331560771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,balanced,0.14086400469144186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,balanced,0.1863306760787964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,balanced,0.23458667596181235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,balanced,0.33243199189503986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,balanced,0.40349332491556805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.06810240149497986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.15322879552841187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.21811840534210206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.2784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.05185279846191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.10177919864654542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.1518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.22979199886322021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.26409599781036375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,power_law_1.2,0.37318398952484133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,balanced,0.033200000723203026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,balanced,0.03245333333810171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,balanced,0.03183466692765554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,balanced,0.040149333576361336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,balanced,0.051327998439470925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,balanced,0.08872000376383464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,balanced,0.10096533099810283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,balanced,0.1288640002409617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,balanced,0.15331199765205383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,balanced,0.21606934070587158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,balanced,0.2661813298861186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,power_law_1.2,0.4919616222381592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,balanced,0.046538665890693665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,balanced,0.04800533254941305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,balanced,0.07148266832033794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,balanced,0.10212799906730652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,balanced,0.1529973347981771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,balanced,0.20341867208480835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,balanced,0.20786132415135702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,balanced,0.20896534125010172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,balanced,0.20697067181269327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,balanced,0.20973332722981772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,balanced,0.2119093338648478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,balanced,0.2160373330116272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,balanced,0.2169333299001058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,balanced,0.22363734245300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,balanced,0.22549333175023398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,balanced,0.23668267329533896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,balanced,0.25154133637746173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,balanced,0.26477332909901935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,balanced,0.2895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,balanced,0.31835732857386273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,balanced,0.36844801902770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,balanced,0.4363413254419963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,balanced,0.5424319903055826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,balanced,0.6721973419189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,balanced,0.904309352238973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,balanced,1.1304372946421306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,8,power_law_1.2,0.21596159934997558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,8,power_law_1.2,0.26141440868377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,8,power_law_1.2,0.281548810005188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,8,power_law_1.2,0.32448639869689944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,8,power_law_1.2,0.3817023992538452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,8,power_law_1.2,0.4373824119567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,8,power_law_1.2,0.40055041313171386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,8,power_law_1.2,0.3641024112701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,8,power_law_1.2,0.3919296026229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,8,power_law_1.2,0.401632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,8,power_law_1.2,0.41194877624511717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,8,power_law_1.2,0.418668794631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,8,power_law_1.2,0.4323391914367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,8,power_law_1.2,0.43928961753845214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,8,power_law_1.2,0.44162559509277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,8,power_law_1.2,0.46352639198303225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,8,power_law_1.2,0.46846718788146974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,8,power_law_1.2,0.508460807800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.2,0.541260814666748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.2,0.6367231845855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.2,0.7347008228302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.2,0.92608642578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.2,1.1104191780090331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.2,1.4747455596923829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.2,2.0272512435913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.2,2.7421119689941404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.2,3.4442111968994142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.18840960264205933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.2747904062271118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.36243839263916017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.2,0.6412735939025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.2,0.6886015892028808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.2,0.7093760013580322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.2,0.7422912120819092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.2,0.7524543762207031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.2,0.7812992095947265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.2,0.8011327743530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.2,0.8394111633300781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.2,0.8586943626403809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.2,0.8964735984802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.2,0.9179455757141113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.2,0.9791808128356934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.101088047027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.2,1.1347328186035157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.2,1.3166655540466308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.2,1.3529279708862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.2,1.6597183227539063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.2,1.9406143188476563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.2,2.5176319122314452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.2,3.1051584243774415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.2,4.344249725341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.2,5.4664958953857425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,balanced,0.06418133278687795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,balanced,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,balanced,0.08660800258318584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,balanced,0.08714666962623596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,balanced,0.08873066306114197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,balanced,0.09104532996813457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,balanced,0.09210667014122009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,balanced,0.09292800227801006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,balanced,0.10094400246938069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,balanced,0.10102400183677673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,balanced,0.10700266559918721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,balanced,0.12223999698956807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.13405866424242655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.1583573321501414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.17225066820780435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.21173334121704102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.25965867439905804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.33799465497334796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.4442773262659709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.6084800163904825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.7917439937591553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,balanced,0.046442667643229164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,balanced,0.05166399975617727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,balanced,0.05199466645717621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,balanced,0.07866133252779643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,balanced,0.08789333701133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,balanced,0.10129599769910176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,balanced,0.11715199549992879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,balanced,0.1564640005429586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,balanced,0.18820265928904215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,balanced,0.25933865706125897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,balanced,0.3306666612625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,power_law_1.2,0.10200320482254029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,power_law_1.2,0.14572160243988036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,power_law_1.2,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,power_law_1.2,0.14497920274734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,power_law_1.2,0.1373952031135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,power_law_1.2,0.133024001121521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,power_law_1.2,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,power_law_1.2,0.13034240007400513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,power_law_1.2,0.14199680089950562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,power_law_1.2,0.1384063959121704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,power_law_1.2,0.1438912034034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,power_law_1.2,0.14112000465393065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,power_law_1.2,0.1349503993988037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,power_law_1.2,0.15267200469970704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,power_law_1.2,0.15719679594039918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,power_law_1.2,0.16840959787368776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,power_law_1.2,0.1772160053253174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,power_law_1.2,0.2019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,power_law_1.2,0.21806719303131103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,power_law_1.2,0.2705215930938721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,power_law_1.2,0.2884864091873169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,power_law_1.2,0.3518336057662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,power_law_1.2,0.39236481189727784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,power_law_1.2,0.5462016105651856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,power_law_1.2,0.6708096027374267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,power_law_1.2,1.0748160362243653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,power_law_1.2,1.3461567878723144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.07789440155029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.07813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.08094720244407654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.08794239759445191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.09753599762916565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.10275839567184449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.1264575958251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.1297727942466736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.15837440490722657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.1614527940750122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.2103935956954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.2512128114700317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,power_law_1.2,0.3554111957550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,power_law_1.2,0.4475264072418213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,power_law_1.2,0.7534527778625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,power_law_1.2,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,power_law_1.2,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,power_law_1.2,0.07820159792900086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,power_law_1.2,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,power_law_1.2,0.08234879970550538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,power_law_1.2,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,power_law_1.2,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,power_law_1.2,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,power_law_1.2,0.12417279481887818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,power_law_1.2,0.13378560543060303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,power_law_1.2,0.9310208320617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,power_law_1.2,0.15927040576934814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,power_law_1.2,0.18681600093841552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,power_law_1.2,0.23152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,power_law_1.2,0.2866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,power_law_1.2,0.3914495944976807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,power_law_1.2,0.4859583854675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,power_law_1.2,0.7296000003814698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,power_law_1.2,0.9485312461853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,balanced,0.02937600016593933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,balanced,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,balanced,0.03617066641648611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,balanced,0.05064000189304352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,balanced,0.06825066606203715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,balanced,0.06883200009663899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,balanced,0.09140266974767049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,balanced,0.09449066718419392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,balanced,0.1329759955406189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,balanced,0.13755200306574503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,balanced,0.05022933085759481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,balanced,0.06564266482988994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,balanced,0.09488000472386678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,balanced,0.1308693289756775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,balanced,0.15929599603017172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,balanced,0.16034133235613504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,balanced,0.16356799999872842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,balanced,0.1625279982884725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,balanced,0.164383997519811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,balanced,0.16429332892100015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,balanced,0.16663466890652975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,balanced,0.170906662940979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,balanced,0.17468800147374472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,balanced,0.18184000253677368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,balanced,0.18614399433135986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,balanced,0.1925333340962728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,balanced,0.21773332357406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,balanced,0.2294879953066508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,balanced,0.2962239980697632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,balanced,0.3210986653963725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,balanced,0.45557868480682373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,balanced,0.4721653461456299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,balanced,0.673472007115682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,balanced,0.7585333188374838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,balanced,1.1141546567281086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,balanced,1.3784106572469075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.073580801486969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.07928959727287292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.1299072027206421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.18325120210647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.2123136043548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.3190783977508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,power_law_1.01,0.4355584144592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,balanced,0.049546668926874794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,balanced,0.05110399921735128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,power_law_1.01,0.5132800102233886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,balanced,0.06046399970849355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,balanced,0.05835199852784475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,balanced,0.06450133522351582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,balanced,0.06649599969387054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,balanced,0.07572799921035767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,balanced,0.08063999811808269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,balanced,0.09312533338864644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,balanced,0.10521599650382996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,balanced,0.11621866623560588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,balanced,0.15075199802716574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,balanced,0.17260799805323282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,balanced,0.22422399123509726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,balanced,0.27110934257507324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,balanced,0.3664160172144572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,balanced,0.4580320119857788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.01,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.01,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.01,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.01,0.07696639895439147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.01,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.01,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.01,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.01,0.09323520064353943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.01,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.01,0.13278080224990846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.01,0.15043840408325196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.01,0.18787839412689208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.01,0.22113919258117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.01,0.281823992729187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.01,0.34523520469665525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.01,0.49083518981933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.01,0.6268415927886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,balanced,0.05160533388455709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,balanced,0.07612266639868419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,balanced,0.10453866918881734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,balanced,0.11917866269747417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,balanced,0.12184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,balanced,0.1257973313331604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,balanced,0.1251573363939921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,balanced,0.12808000048001608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,balanced,0.1241386632124583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,balanced,0.1260373294353485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,balanced,0.12818666299184164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,balanced,0.13085333506266275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,balanced,0.13788800438245138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,balanced,0.14174399773279825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,balanced,0.14936000108718872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,balanced,0.166512002547582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,balanced,0.17818133036295572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,balanced,0.20962133010228476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,balanced,0.2353973388671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,balanced,0.3123626708984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,balanced,0.35331201553344727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,balanced,0.49380799134572345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,balanced,0.6055253346761068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,balanced,0.83570663134257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,balanced,1.0967360337575276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,balanced,0.07841599980990092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,balanced,0.13314132889111838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,balanced,0.23344000180562338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,balanced,0.23702933390935263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,balanced,0.23651733001073202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,balanced,0.2359093427658081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,balanced,0.23747199773788452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,balanced,0.239029328028361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,balanced,0.23812800645828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,balanced,0.23990933100382486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,balanced,0.24338134129842123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,balanced,0.24412800868352255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,balanced,0.24839999278386435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,balanced,0.2568639914194743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,balanced,0.26254934072494507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,balanced,0.26737066109975177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,balanced,0.29341334104537964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,balanced,0.3163253267606099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,balanced,0.3457119862238566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,balanced,0.4315733512242635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,balanced,0.48002131779988605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,balanced,0.6566666762034098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,balanced,0.7952799797058105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,balanced,1.161626656850179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,balanced,1.5183253288269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.06773759722709656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,balanced,2.2246665954589844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.09425920248031616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,balanced,2.9306774139404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.11312639713287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.13019520044326782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.1556607961654663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.16005120277404786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.15916160345077515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.16476800441741943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.17493120431900025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.1722815990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.1786239981651306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.1946303963661194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.20257918834686278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.20687360763549806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.22691841125488282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.2571903944015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.29859840869903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.3478463888168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.45563521385192873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.549945592880249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.7467328071594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.9508607864379883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.3346495628356934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,1.7885503768920898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,power_law_1.01,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,power_law_1.01,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,power_law_1.01,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,power_law_1.01,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,power_law_1.01,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,power_law_1.01,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,power_law_1.01,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,power_law_1.01,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,power_law_1.01,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,power_law_1.01,0.14590079784393312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,power_law_1.01,0.17111680507659913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,power_law_1.01,0.22673919200897216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,power_law_1.01,0.3705471992492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,power_law_1.01,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,power_law_1.01,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,power_law_1.01,0.08693119883537292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,power_law_1.01,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,power_law_1.01,0.11205120086669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,power_law_1.01,0.11797759532928467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,power_law_1.01,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,power_law_1.01,0.11861759424209595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,power_law_1.01,0.12327040433883667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,power_law_1.01,0.1299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,power_law_1.01,0.12962559461593628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,power_law_1.01,0.13607679605484008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,power_law_1.01,0.13995519876480103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,power_law_1.01,0.14349440336227418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,power_law_1.01,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,power_law_1.01,0.1725119948387146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.01,0.187225604057312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.01,0.22595839500427245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.01,0.2662911891937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.01,0.33990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.01,0.4112895965576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.01,0.5873792171478271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.01,0.7222976207733154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.01,1.0532928466796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.01,1.3897791862487794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,power_law_1.01,0.4664447784423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,power_law_1.01,0.5640768051147461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,power_law_1.2,0.027449598908424376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,power_law_1.2,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,power_law_1.2,0.027270400524139406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,power_law_1.2,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,power_law_1.2,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,power_law_1.2,0.028697600960731505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,power_law_1.2,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,power_law_1.2,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,power_law_1.2,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,power_law_1.2,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,power_law_1.2,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,power_law_1.2,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,power_law_1.2,0.1174847960472107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,power_law_1.2,0.16429439783096314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,power_law_1.2,0.22540159225463868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,power_law_1.2,0.28764159679412843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.18424320220947266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.2256704092025757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.31527678966522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.040505599975585935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.0573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.059520000219345094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,power_law_1.01,0.045203199982643126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,power_law_1.01,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,power_law_1.01,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,power_law_1.01,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.10023679733276367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,power_law_1.01,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,power_law_1.01,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,power_law_1.01,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,power_law_1.01,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,power_law_1.01,0.09005439877510071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,power_law_1.01,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.12310400009155273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.14713599681854247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,power_law_1.01,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.20521600246429444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,power_law_1.01,0.19396480321884155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.22639360427856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,power_law_1.01,0.2583168029785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.33512320518493655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,power_law_1.01,0.3469183921813965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.4165823936462402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,power_law_1.01,0.44800639152526855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.6140223979949951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.7639232158660889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.3661056041717529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,power_law_1.01,0.024108800292015075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,power_law_1.01,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,power_law_1.01,0.027532801032066345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,power_law_1.01,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,power_law_1.01,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,power_law_1.01,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,power_law_1.01,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,power_law_1.01,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,power_law_1.01,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,power_law_1.01,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.11385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.1235967993736267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.14991999864578248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.19498879909515382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.27850239276885985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.3456255912780762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.4937856197357178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,power_law_1.01,0.6556352138519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.09422079920768738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.12053120136260986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.1538432002067566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.1912511944770813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.19774719476699829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.2722048044204712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.2690239906311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.26104960441589353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.28433918952941895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.2889343976974487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.3007551908493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.3096191883087158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.3217344045639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.34639999866485593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.3661887884140015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.39239039421081545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.42459521293640134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.4960000038146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.5379712104797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.6815616130828858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,power_law_1.2,0.8233983993530274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.104742431640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.3362048149108887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,power_law_1.2,1.8304384231567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,power_law_1.2,2.0681663513183595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,balanced,0.05936000247796377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,balanced,0.05845866600672404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,balanced,0.06654933094978333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,balanced,0.11939199765523274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,balanced,0.15010133385658264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,balanced,0.1483519971370697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,balanced,0.14731199542681375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,balanced,0.14731199542681375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,balanced,0.15011200308799744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,balanced,0.1521013379096985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,balanced,0.15427199999491373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,balanced,0.15548800428708395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,balanced,0.16379200418790182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,balanced,0.16402666767438254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,balanced,0.17108800013860068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,balanced,0.186191995938619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.19898132483164468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.22631466388702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.2524906595547994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.30661867062250775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.36418668429056805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.48284268379211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.63264532883962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.8694026470184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,balanced,1.1469439665476482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,balanced,0.030074665943781536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,balanced,0.03012799968322118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,balanced,0.029290666182835896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,balanced,0.03257066756486893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,balanced,0.034485332667827606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,balanced,0.06966933111349742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,balanced,0.09153067072232564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,balanced,0.12145599722862244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,balanced,0.14848533272743225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.2,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.2,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.2,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.2,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.2,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.2,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.2,0.0954367995262146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.2,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.2,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.2,0.14051200151443483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.2,0.13203200101852416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.2,0.16428799629211427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.2,0.16782079935073851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.2,0.2412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.2,0.3367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.2,0.4401408195495605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.2,0.4310272216796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,power_law_1.2,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,power_law_1.2,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,power_law_1.2,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,power_law_1.2,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,power_law_1.2,0.08820480108261108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,power_law_1.2,0.10299520492553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,power_law_1.2,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,power_law_1.2,0.13957120180130006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,power_law_1.2,0.17367680072784425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,power_law_1.2,0.22016000747680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,power_law_1.2,0.2847615957260132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,power_law_1.2,0.38317439556121824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,power_law_1.2,0.5035200119018555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,power_law_1.2,0.7127679824829102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,power_law_1.2,0.9228927612304687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,balanced,0.05460800230503082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,balanced,0.05645333230495453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,balanced,0.05522666871547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,balanced,0.05458133419354757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,balanced,0.05781333148479462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,balanced,0.060346667965253196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,balanced,0.06272533535957336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,balanced,0.06528533498446147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,balanced,0.07131200035413106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,balanced,0.08218666911125183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,balanced,0.08861866593360901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,balanced,0.10149866342544556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,balanced,0.12723199526468912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,balanced,0.15480533242225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,balanced,0.18401066462198892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,power_law_1.2,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,power_law_1.2,0.08702080249786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,power_law_1.2,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,power_law_1.2,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,power_law_1.2,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,power_law_1.2,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,power_law_1.2,0.10248960256576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,power_law_1.2,0.10668799877166749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,power_law_1.2,0.11738239526748658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,power_law_1.2,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,power_law_1.2,0.14106240272521972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,power_law_1.2,0.1789888024330139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,power_law_1.2,0.19612799882888793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,power_law_1.2,0.25080320835113523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,power_law_1.2,0.31633920669555665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,power_law_1.2,0.4488192081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,power_law_1.2,0.5593920230865479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,power_law_1.2,0.7799359798431397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,power_law_1.2,1.090931224822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,power_law_1.01,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,power_law_1.01,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,power_law_1.01,0.12194559574127198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,power_law_1.01,0.11787519454956055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,power_law_1.01,0.1242624044418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,power_law_1.01,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,power_law_1.01,0.12838399410247803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,power_law_1.01,0.12991360425949097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,power_law_1.01,0.13436800241470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,power_law_1.01,0.13341439962387086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,power_law_1.01,0.14638079404830934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,power_law_1.01,0.15756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,power_law_1.01,0.175545597076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,power_law_1.01,0.19147520065307616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,power_law_1.01,0.2249664068222046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,power_law_1.01,0.25322880744934084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,power_law_1.01,0.3052736043930054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,power_law_1.01,0.3691135883331299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,power_law_1.01,0.48264322280883787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,power_law_1.01,0.6784832000732421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,power_law_1.01,0.8893888473510743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,power_law_1.01,1.2801728248596191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.12035200595855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.17282559871673583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.23578240871429443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.23814399242401124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.24035840034484862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.2452415943145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.24759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.24921600818634032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.25061759948730467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.25867519378662107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.26328320503234864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.2684927940368652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.280134391784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.30631680488586427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.31597440242767333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.3425535917282104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.38382720947265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.45127038955688475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.557862377166748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,power_law_1.01,1.7258304595947265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.6752319812774659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,0.8774592399597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,1.0977984428405763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,1.538323211669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,1.984294319152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,2.8519935607910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,power_law_1.01,1.9156671524047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,3.7211326599121093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,power_law_1.01,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,power_law_1.01,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,power_law_1.01,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,power_law_1.01,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,power_law_1.01,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,power_law_1.01,0.061536002159118655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,power_law_1.01,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,power_law_1.01,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,power_law_1.01,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,power_law_1.01,0.111571204662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,power_law_1.01,0.12577279806137084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,power_law_1.01,0.1503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,power_law_1.01,0.17383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,power_law_1.01,0.23573760986328124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,power_law_1.01,0.2938944101333618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,power_law_1.01,0.446668815612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,power_law_1.01,0.5478591918945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,power_law_1.2,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,power_law_1.2,0.06147199869155884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.2,0.16178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.2,0.19286400079727173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,balanced,0.07348800202210744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,balanced,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,balanced,0.12602667013804117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,balanced,0.12663466731707254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,balanced,0.12685867150624594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,balanced,0.12614933649698892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,balanced,0.12970133622487387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,balanced,0.12990933656692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,balanced,0.13036266962687174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,balanced,0.134250670671463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,balanced,0.13876799742380777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,balanced,0.1420693298180898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,balanced,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,balanced,0.15731733043988547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,balanced,0.16570132970809937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,balanced,0.18783466021219888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,balanced,0.20710933208465576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,balanced,0.24436267217000326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,balanced,0.28723732630411786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,balanced,0.36057066917419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,balanced,0.4427306652069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,balanced,0.5948906739552816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,balanced,0.768986701965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.2,0.27608959674835204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.2,0.35564799308776857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.2,0.551251220703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,balanced,0.029450667401154835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,balanced,0.02938666691382726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,balanced,0.031744000812371574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,balanced,0.03384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,balanced,0.06622399886449178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,balanced,0.09663466612497966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,balanced,0.11769599715868632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,balanced,0.1646880010763804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,balanced,0.20551466941833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.2,0.7721920013427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.2,1.0798080444335938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,balanced,0.04208533465862274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,balanced,0.09063466389973958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,balanced,0.11346133550008138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,balanced,0.11598933736483256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,balanced,0.117658664782842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,balanced,0.1172160009543101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,balanced,0.1197760005791982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,balanced,0.1204213301340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,balanced,0.12257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,balanced,0.1311840017636617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,balanced,0.1320373316605886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,balanced,0.13460266590118408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,balanced,0.14813866217931113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,balanced,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,balanced,0.18149334192276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,balanced,0.20648000637690225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,balanced,0.25882667303085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,balanced,0.32150399684906006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,balanced,0.4397226572036743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,balanced,0.5452586809794108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,balanced,0.7971200148264567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,balanced,1.0144426822662354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.2,1.4963392257690429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,balanced,0.06435200075308482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,balanced,0.09146666526794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,balanced,0.1104853351910909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,balanced,0.11186666289965312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,balanced,0.10672000050544739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,balanced,0.10593066612879436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,balanced,0.10616532961527507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,balanced,0.10753599802652995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,balanced,0.11004799604415894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,balanced,0.11319999893506368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,balanced,0.12034133076667786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,balanced,0.1220266620318095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,balanced,0.132560004790624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,balanced,0.15440000096956888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,balanced,0.16050133109092712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,balanced,0.22451732556025186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,balanced,0.2452639937400818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,balanced,0.3583039840062459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,balanced,0.38894399007161456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,balanced,0.5633226633071899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,balanced,0.6816426912943522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,balanced,1.029077370961507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,balanced,1.2640693187713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.1185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.11418240070343018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.09642239809036254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.09784960150718688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.10865919589996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.11415679454803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.1247871994972229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.13062399625778198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.1368191957473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.13699840307235717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.1502079963684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.17840640544891356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.20817921161651612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.25839359760284425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.2826047897338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,power_law_1.2,0.37240960597991946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,power_law_1.2,0.3868607997894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,power_law_1.2,0.5783552169799805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,power_law_1.2,0.7439487934112549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,power_law_1.2,0.950649642944336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,power_law_1.2,1.415993595123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,power_law_1.01,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,power_law_1.01,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,power_law_1.01,0.04294399917125702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,power_law_1.01,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,power_law_1.01,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,power_law_1.01,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,power_law_1.01,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,power_law_1.01,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,power_law_1.01,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,power_law_1.01,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,power_law_1.01,0.15040639638900757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,power_law_1.01,0.1723647952079773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,power_law_1.01,0.2328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,power_law_1.01,0.2933120012283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,power_law_1.01,0.4810880184173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,power_law_1.01,0.6920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.08021759986877441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.08536959886550903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.09478399753570557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.13186559677124024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.136627197265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.14149760007858275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,balanced,0.04349866509437561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,balanced,0.04608533283074697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,balanced,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,balanced,0.08252266546090443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,balanced,0.09925333658854167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.13963520526885986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,balanced,0.11905066172281902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,balanced,0.17454399665196738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,balanced,0.21572266022364298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,balanced,0.30110933383305866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.14588799476623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,balanced,0.389024019241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.15315200090408326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.15805439949035643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.15786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.16136319637298585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.17811839580535888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.18252160549163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.20932478904724122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.23576319217681885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.25306239128112795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,power_law_1.2,0.3607232093811035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,power_law_1.2,0.37256319522857667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,power_law_1.2,0.4957183837890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,power_law_1.2,0.6247104167938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,power_law_1.2,0.8746623992919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,balanced,0.06782400111357371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,balanced,0.09062400460243225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,balanced,0.12494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,balanced,0.19771732886632284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,balanced,0.329258660475413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,balanced,0.45767998695373535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,balanced,0.4607786734898885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,balanced,0.46029333273569745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,balanced,0.46107200781504315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,balanced,0.46484800179799396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,balanced,0.4689759810765584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,balanced,0.4726239840189616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,balanced,0.47786132494608563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,balanced,0.4809226592381795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,balanced,0.493008017539978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,balanced,0.49604801336924237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,balanced,0.5126986503601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,balanced,0.5475680033365885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,balanced,0.5725813309351603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,balanced,0.7408746878306071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,balanced,0.6946826775868734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,balanced,1.0135893026987712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,balanced,0.9518453280131022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,balanced,1.4010826746622722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,balanced,1.546346664428711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,balanced,2.3354293505350747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,balanced,2.6855252583821616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,power_law_1.2,1.1362175941467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,power_law_1.2,1.5030976295471192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.01,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.01,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.01,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.01,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.01,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.01,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.01,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.01,0.09329280257225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.01,0.09937919974327088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.01,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.01,0.11939840316772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.01,0.14793599843978883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.01,0.15928959846496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.01,0.20455679893493653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.01,0.23301761150360106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.01,0.3031487941741943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.01,0.32448639869689944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.01,0.5162816047668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.01,0.5471807956695557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,power_law_1.2,2.2387136459350585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.10179840326309204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.12475520372390747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.12556159496307373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.1404031991958618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.15057920217514037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.16470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.16736639738082887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.16876800060272218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.16753920316696166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.17419519424438476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.17905919551849364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.18301440477371217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.18643840551376342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.1972991943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.21196160316467286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.20769920349121093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.25615999698638914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.2495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.31633920669555665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.31176319122314455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.4342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,power_law_1.2,0.42653441429138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,power_law_1.2,0.5365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,power_law_1.2,0.7340799808502197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,power_law_1.2,1.019814395904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,power_law_1.2,1.4228223800659179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,balanced,0.050554667909940086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,balanced,0.05203733344872793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,balanced,0.0543093333641688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,balanced,0.0721973329782486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,balanced,0.07197333375612895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,balanced,0.07067200044790904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,balanced,0.07286933561166127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,balanced,0.0874079962571462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,balanced,0.09285866220792134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,balanced,0.10707733035087585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,balanced,0.1102133293946584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,balanced,0.13372266292572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,balanced,0.1464853286743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,balanced,0.18525334199269614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,balanced,0.21213332811991373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,balanced,0.27115732431411743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,balanced,0.33980798721313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,balanced,0.4679946502049764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,balanced,0.6056373516718546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.0449535995721817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.05969280004501343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.12696319818496704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.1719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.19406720399856567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.3310271978378296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.40018558502197266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,balanced,0.027482666075229645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,balanced,0.03180266668399175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,balanced,0.031157332162062328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,balanced,0.03314133236805598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,balanced,0.04404266675313314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,balanced,0.082805335521698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,balanced,0.09964799880981445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,balanced,0.1322933336098989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,balanced,0.16182933251063028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,power_law_1.2,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,power_law_1.2,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,power_law_1.2,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,power_law_1.2,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,power_law_1.2,0.05013759732246399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,power_law_1.2,0.05185279846191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,power_law_1.2,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,power_law_1.2,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,power_law_1.2,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.2,0.11543680429458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.2,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.2,0.19695359468460083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.2,0.25173120498657225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.2,0.35489280223846437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.2,0.460211181640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.01,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.01,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.01,0.10928000211715698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.01,0.12078720331192017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.01,0.15104000568389891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.01,0.1772480010986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.2,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.2,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.2,0.15134719610214234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.2,0.17072000503540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.2,0.23994879722595214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.2,0.21825919151306153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.2,0.23226239681243896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.2,0.22894721031188964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.2,0.22607998847961425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.2,0.2392319917678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.2,0.24249598979949952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.2,0.25473918914794924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.2,0.25768959522247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.2,0.2710911989212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.2,0.2739583969116211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.2,0.29793920516967776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.2,0.3408384084701538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.2,0.35445120334625246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.2,0.42397441864013674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.2,0.4816768169403076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.2,0.6061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.2,0.6656447887420655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.2,0.8514944076538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.2,1.120531177520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.2,1.5733247756958009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.2,2.0015039443969727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,balanced,0.05329066514968872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,balanced,0.07792533437410991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,balanced,0.0841919978459676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,balanced,0.08532266815503438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,balanced,0.0863200028737386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,balanced,0.08563199639320374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,balanced,0.08767466743787129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,balanced,0.08705600102742513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,balanced,0.08989866574605306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,balanced,0.09143466750780742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,balanced,0.09643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,balanced,0.10065066814422607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,balanced,0.11058666308720906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,balanced,0.11629866560300191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,balanced,0.136245330174764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,balanced,0.15214932958285013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,balanced,0.19340266784032187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,balanced,0.21358400583267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,balanced,0.28959999481836957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,balanced,0.3516639868418376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,balanced,0.4842453400293986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,balanced,0.6183893283208212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,balanced,0.030063999195893604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,balanced,0.029663999875386555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,balanced,0.03316266586383184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,balanced,0.03647999962170919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,balanced,0.036837334434191384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,balanced,0.04613866905371348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,balanced,0.0503359983364741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,balanced,0.060378665725390114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,balanced,0.0776693324247996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,balanced,0.11770666639010112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,balanced,0.14150399963061014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,balanced,0.1953173279762268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,balanced,0.24125333627065024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,balanced,0.031930667658646904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,balanced,0.03496533383925756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,balanced,0.03594133257865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,balanced,0.036101333796978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,balanced,0.04543466866016388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,balanced,0.05367999772230784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,balanced,0.0707946668068568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,balanced,0.09077866872151692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,balanced,0.13035733501116434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,balanced,0.15998933712641397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.08894079923629761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09082239866256714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.10509439706802368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11530879735946656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.12307840585708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.1275904059410095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.13322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.16773760318756104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.18243199586868286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.23442559242248534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.297273588180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.420147180557251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.5320703983306885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.7474048137664795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,power_law_1.2,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,power_law_1.2,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,power_law_1.2,0.059520000219345094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,power_law_1.2,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,power_law_1.2,0.06504319906234741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,power_law_1.2,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,power_law_1.2,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,power_law_1.2,0.11163519620895386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.9706687927246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,power_law_1.2,0.15285760164260864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,power_law_1.2,0.17415039539337157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.405235195159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.06909440159797668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,power_law_1.2,0.22545280456542968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.12772480249404908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.1562432050704956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,power_law_1.2,0.25664639472961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.16998399496078492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.2304896116256714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.28453760147094725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.8445375442504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,power_law_1.2,0.3837824106216431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.37463040351867677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.47122559547424314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,power_law_1.2,0.5181503772735596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.7524415969848632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,power_law_1.2,0.7739456176757813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.8118144035339355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,power_law_1.2,0.9102911949157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,power_law_1.2,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,power_law_1.2,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,power_law_1.2,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,power_law_1.2,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,power_law_1.2,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,power_law_1.2,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,power_law_1.2,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,power_law_1.2,0.09580159783363343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,power_law_1.2,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.2,0.1132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.2,0.1403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.2,0.1651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.2,0.2047935962677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.2,0.25428481101989747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.2,0.35662078857421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.2,0.47223677635192873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.2,0.7236032009124755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.2,0.9651840209960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,balanced,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,balanced,0.127920001745224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,balanced,0.18666134277979532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,balanced,0.3111520012219747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,balanced,0.5510026613871256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,balanced,1.0275039672851562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,balanced,1.034325361251831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,balanced,1.0384853680928547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,balanced,1.0429600079854329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,balanced,1.043946663538615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,balanced,1.0494506359100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,balanced,1.0572693347930908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,balanced,1.0667733351389568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,balanced,1.0723093350728352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,balanced,1.0848159790039062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,balanced,1.0936426321665447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,balanced,1.1210239728291829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,balanced,1.159157355626424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.1941546599070232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.2794293562571208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.3877493540445964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.5417173703511555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.7121119499206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,balanced,2.320927937825521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,balanced,2.5560213724772134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,balanced,3.702672004699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,balanced,4.584351857503255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,power_law_1.2,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,power_law_1.2,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,power_law_1.2,0.2002432107925415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,power_law_1.2,0.23621759414672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,power_law_1.2,0.33736319541931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,power_law_1.2,0.40010881423950195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,power_law_1.2,0.41220479011535643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,power_law_1.2,0.5017407894134521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,power_law_1.2,0.5079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,power_law_1.2,0.5465023994445801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,power_law_1.2,0.5221375942230224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,power_law_1.2,0.5659327983856202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,power_law_1.2,0.5348351955413818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,power_law_1.2,0.5402304172515869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,power_law_1.2,0.5768896102905273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,power_law_1.2,0.5827328205108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,power_law_1.2,0.6024896144866944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,power_law_1.2,0.6430848121643067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,power_law_1.2,0.7056384086608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,power_law_1.2,0.8343935966491699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,power_law_1.2,0.8315839767456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,power_law_1.2,1.014195156097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,power_law_1.2,1.178438377380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,power_law_1.2,1.511513614654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,power_law_1.2,1.6356416702270509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,power_law_1.2,2.080339241027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,power_law_1.2,2.513350486755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,balanced,0.059994667768478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,balanced,0.06042666733264923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,balanced,0.059893334905306496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,balanced,0.06011733412742615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,balanced,0.060453335444132485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,balanced,0.06815466781457265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,balanced,0.07090133428573608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,balanced,0.08073066671689351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,balanced,0.08725333213806152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,balanced,0.11198932925860088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,balanced,0.12443733215332031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,balanced,0.1710240046183268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,balanced,0.16479466358820596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,balanced,0.20808533827463785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,balanced,0.2329066594441732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,balanced,0.30459733804066974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,balanced,0.3643360137939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,balanced,0.031514666974544525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,balanced,0.03346666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,balanced,0.033957332372665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,balanced,0.03605866680542628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,balanced,0.036090667049090065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,balanced,0.06163733204205831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,balanced,0.07418133318424225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,balanced,0.07627733548482259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,balanced,0.09798933068911235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,balanced,0.10714667042096455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,balanced,0.14713600277900696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,balanced,0.15826666355133057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,balanced,0.22320000330607095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,balanced,0.2829759915669759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,balanced,0.401253342628479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,balanced,0.5060586531956991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.2,0.07326719760894776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.2,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.2,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.2,0.10208640098571778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.2,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.2,0.10087039470672607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.2,0.1029312014579773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.2,0.10550400018692016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.2,0.1085312008857727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.2,0.11292159557342529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.2,0.12079999446868897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.2,0.12567039728164672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.2,0.14287359714508058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.2,0.16436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.2,0.17037440538406373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.2,0.20485119819641112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.2,0.25296640396118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.2,0.33323519229888915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.2,0.3891072034835815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.2,0.5086143970489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.2,0.6486911773681641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.2,0.8246975898742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.2,1.0849984169006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.14961279630661012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.17571200132369996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.2622143983840942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.33182721138000487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.3694272041320801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.47303037643432616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.4545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.5047743797302247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.5111551761627198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.5249343872070312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.5375999927520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.5370816230773926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.5752575874328614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.5697728157043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.6078080177307129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.6311552047729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.6879487991333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,power_law_1.01,0.7944128036499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,power_law_1.01,0.8110207557678223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,power_law_1.01,0.9469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.084883213043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,power_law_1.01,1.3576000213623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,power_law_1.01,1.5303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,power_law_1.01,2.040390396118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,power_law_1.01,2.4471551895141603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,power_law_1.2,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,power_law_1.2,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,power_law_1.2,0.07296640276908875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,power_law_1.2,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,power_law_1.2,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,power_law_1.2,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,power_law_1.2,0.10403200387954711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,power_law_1.2,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,power_law_1.2,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,power_law_1.2,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,power_law_1.2,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,power_law_1.2,0.11927679777145386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,power_law_1.2,0.12879999876022338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,power_law_1.2,0.13262079954147338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,power_law_1.2,0.1463104009628296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,power_law_1.2,0.15999360084533693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.2,0.1765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.2,0.2153088092803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.2,0.2515775918960571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.2,0.3191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.2,0.40142078399658204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.2,0.5517695903778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.2,0.6825088024139404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.2,0.9768959999084472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.2,1.3133695602416993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,power_law_1.2,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,power_law_1.2,0.09201920032501221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,power_law_1.2,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,power_law_1.2,0.12162560224533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,power_law_1.2,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,power_law_1.2,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,power_law_1.2,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,power_law_1.2,0.13188480138778685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,power_law_1.2,0.13021440505981446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,power_law_1.2,0.13338240385055541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,power_law_1.2,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,power_law_1.2,0.14351999759674072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,power_law_1.2,0.1490623950958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,power_law_1.2,0.16017279624938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,power_law_1.2,0.16634880304336547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,power_law_1.2,0.19058560132980346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,power_law_1.2,0.21712000370025636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,power_law_1.2,0.25759999752044677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,power_law_1.2,0.32126080989837646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,power_law_1.2,0.38996479511260984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,power_law_1.2,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,power_law_1.2,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,power_law_1.2,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,power_law_1.2,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,power_law_1.2,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,power_law_1.2,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,power_law_1.2,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,power_law_1.2,0.46403841972351073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,power_law_1.2,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,power_law_1.2,0.6250432014465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,power_law_1.2,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,power_law_1.2,0.12220799922943115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,power_law_1.2,0.14710400104522706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,power_law_1.2,0.19006719589233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.09569919705390931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.11724799871444702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.16406400203704835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.17702399492263793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.18369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.18933119773864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.19571199417114257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.19827840328216553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.2106175899505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.2170624017715454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.22224640846252441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.2368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2454911947250366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.26469759941101073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.3065023899078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.3471872091293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.42369279861450193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.49722881317138673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6531712055206299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8435135841369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.111616039276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,power_law_1.2,0.95283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.5180543899536132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.019968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.505580711364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,power_law_1.2,0.2273087978363037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,power_law_1.2,1.101318359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,power_law_1.2,0.3172607898712158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.2,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.2,0.10366719961166382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.2,0.12175359725952148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.2,0.15290240049362183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.2,0.1756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.2,0.23106560707092286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.2,0.2284032106399536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.2,0.2368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.2,0.2336127996444702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.2,0.2313215970993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.2,0.24299519062042235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.2,0.25166079998016355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.2,0.25036160945892333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.2,0.2687616109848022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.2,0.28672640323638915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.2,0.2941375970840454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,power_law_1.2,0.4087808132171631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.2,0.28923521041870115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.2,0.3307071924209595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.2,0.35371520519256594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.2,0.4214208126068115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.2,0.47617278099060056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.2,0.6114943981170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.2,0.6730368137359619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.2,0.8868672370910644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.2,1.1015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.2,1.5886336326599122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.2,1.9232255935668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,power_law_1.2,1.5739392280578612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,power_law_1.2,0.5802303791046143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,power_law_1.2,2.081760025024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,power_law_1.2,0.7831232070922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,power_law_1.01,0.12079999446868897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,power_law_1.01,0.18090879917144775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,power_law_1.01,0.23197441101074218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,power_law_1.01,0.3743423938751221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,power_law_1.01,0.3729151964187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,power_law_1.01,0.3872319936752319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,power_law_1.01,0.39479680061340333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,power_law_1.01,0.39183359146118163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,power_law_1.01,0.4074240207672119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,power_law_1.01,0.3962687969207764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,power_law_1.01,0.42785282135009767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,power_law_1.01,0.4437056064605713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,power_law_1.01,0.4416639804840088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,power_law_1.01,0.4493055820465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,power_law_1.01,0.48293118476867675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,power_law_1.01,0.48687357902526857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,power_law_1.01,0.5155007839202881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,power_law_1.01,0.6140927791595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,power_law_1.01,0.6827328205108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,power_law_1.01,0.8515007972717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,power_law_1.01,0.9877440452575683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,power_law_1.01,1.3193728446960449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,power_law_1.01,1.6332351684570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.06959999799728393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,power_law_1.01,2.3083072662353517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.11738879680633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,balanced,0.05490666627883911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,balanced,0.0632533331712087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,balanced,0.0820000022649765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,balanced,0.11429333686828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,balanced,0.13126400113105774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,balanced,0.13475199540456137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,balanced,0.13731200496355692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,balanced,0.13730133573214212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,balanced,0.1406826674938202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,balanced,0.13692266742388406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,balanced,0.13823466499646506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,balanced,0.14204266667366028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,balanced,0.1439626713593801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,balanced,0.15100266536076865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,balanced,0.15550933281580606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,balanced,0.1621280014514923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,balanced,0.18338133891423544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,power_law_1.01,2.9947071075439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,balanced,0.20253866910934448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,balanced,0.2433333396911621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,balanced,0.2731893261273702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,balanced,0.3941333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,balanced,0.46161067485809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,balanced,0.6696693102518717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,balanced,0.8285333315531412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,balanced,1.211568037668864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,balanced,1.5511520703633626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.1725376009941101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,power_law_1.01,4.362252807617187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.2122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,power_law_1.01,5.780275344848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.30550398826599123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.4438015937805176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.11213439702987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.11945600509643554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.12686079740524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.12960000038146974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.13299839496612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.14416639804840087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.14925440549850463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.15845760107040405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.16486400365829468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.1807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.2069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.2300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.28054399490356446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3208640098571777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.4242368221282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.5160192012786865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.7079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,power_law_1.01,0.8869503974914551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.2630528450012206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.665567970275879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.061939197778701785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.11329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.11637760400772094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.15716480016708373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.1863168001174927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.2766144037246704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.34335999488830565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.4882815837860107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.09271039962768554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.1094655990600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.1359488010406494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.13576960563659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.14086400270462035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.14877439737319947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.17118079662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.18602880239486694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.19740159511566163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.15624959468841554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.16286720037460328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.16215039491653443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.1880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.20049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.21791999340057372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.23281919956207275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.27278718948364256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,power_law_1.2,0.35656960010528566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,power_law_1.2,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,power_law_1.2,0.46078081130981446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,power_law_1.2,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,power_law_1.2,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,power_law_1.2,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,power_law_1.2,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,power_law_1.2,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,power_law_1.2,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,power_law_1.2,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,power_law_1.2,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,power_law_1.2,0.11363199949264527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.2,0.12378239631652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,power_law_1.2,0.5376383781433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.2,0.14973440170288085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.2,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.2,0.2239487886428833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.2,0.273907208442688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.2,0.3788800001144409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.2,0.4922048091888428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.2,0.7064767837524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.2,0.9634495735168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,power_law_1.2,0.6944191932678223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.6467008113861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,power_law_1.2,0.9518527984619141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,power_law_1.2,1.3558655738830567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,power_law_1.2,1.6556415557861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,power_law_1.2,2.4307647705078126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,power_law_1.2,3.399385452270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.12400640249252319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.13141759634017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.1342463970184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.13352960348129272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.1371072053909302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.1415488004684448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.14658559560775758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.1546623945236206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.1650879979133606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.17237759828567506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.18955520391464234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.22505600452423097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,balanced,0.07749866445859273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,balanced,0.07800533374150594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.25287680625915526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,balanced,0.07930133243401845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.3202816009521484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,balanced,0.08155733346939087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.3790015935897827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,balanced,0.09633599718411763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,balanced,0.10817066828409831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,balanced,0.11542933185895284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,balanced,0.13248533010482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.516761589050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,balanced,0.16101866960525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,balanced,0.19511467218399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,balanced,0.22737600406010947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,balanced,0.3032426635424296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,balanced,0.3762880166371663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,balanced,0.5316960016886393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,balanced,0.6824266910552979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.6429567813873291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.9064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.1574015617370605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,1.671232032775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,power_law_1.01,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,power_law_1.01,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,power_law_1.01,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,power_law_1.01,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,power_law_1.01,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,2.157548713684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,power_law_1.01,0.07563520073890687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,power_law_1.01,0.0822655975818634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,power_law_1.01,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,power_law_1.01,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,power_law_1.01,0.16012799739837646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,power_law_1.01,0.2016511917114258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,power_law_1.01,0.26207358837127687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,power_law_1.01,0.3361855983734131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.07614719867706299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.0953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.11944320201873779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.15927679538726808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.1789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.2812096118927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.35111041069030763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.5144576072692871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.6657087802886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.9956416130065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,1.297651195526123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,power_law_1.01,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,power_law_1.01,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,power_law_1.01,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,power_law_1.01,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,power_law_1.01,0.09893760085105896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,power_law_1.01,0.11745920181274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,power_law_1.01,0.12191359996795655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,power_law_1.01,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,power_law_1.01,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,power_law_1.01,0.1342144012451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,power_law_1.01,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,balanced,0.023472001155217487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,balanced,0.02532266577084859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,balanced,0.023370665808518726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,power_law_1.01,0.14714879989624025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,balanced,0.023706667125225067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,balanced,0.02536533276240031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,balanced,0.026496000587940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,balanced,0.027290667096773785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,balanced,0.025285333395004272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,power_law_1.01,0.15835520029067993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,balanced,0.025685332715511322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,balanced,0.027285332481066387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,balanced,0.02749866743882497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,balanced,0.02755733331044515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,balanced,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,power_law_1.01,0.1659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,power_law_1.01,0.1761728048324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,power_law_1.01,0.1792896032333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,power_law_1.01,0.21155838966369628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,power_law_1.01,0.2347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,power_law_1.01,0.2964544057846069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,power_law_1.01,0.35471999645233154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,power_law_1.01,0.463756799697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,power_law_1.01,0.5719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,power_law_1.01,0.791270399093628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,power_law_1.01,1.0165056228637694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,balanced,0.03514666606982549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,power_law_1.01,1.4560704231262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,balanced,0.05235200126965841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,balanced,0.06089599927266439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,power_law_1.01,1.8928319931030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,balanced,0.05818133552869161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,balanced,0.10070400436719258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,balanced,0.15482133626937866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,balanced,0.2600853244463603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,balanced,0.4578506549199422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,balanced,0.6558880011240641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,balanced,0.6675679683685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,balanced,0.6767679850260416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,balanced,0.6721866925557455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,balanced,0.6779733498891195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,balanced,0.6822613080342611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,balanced,0.6853813330332438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,balanced,0.7017760276794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,balanced,0.7054986953735352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,balanced,0.7167519728342692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,balanced,0.7410613695780436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,balanced,0.7853279908498129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,balanced,0.8200053373972574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,balanced,0.9011306762695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,balanced,0.9775839646657308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,balanced,1.1417653560638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,balanced,1.337733268737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,balanced,1.6620906194051106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,balanced,2.11788272857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,balanced,2.752842585245768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,balanced,3.759903907775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,power_law_1.01,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,power_law_1.01,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,power_law_1.01,0.04190079867839813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,power_law_1.01,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,power_law_1.01,0.0522816002368927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,power_law_1.01,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.01,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.01,0.1272447943687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.01,0.15438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.01,0.21532158851623534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.01,0.28550400733947756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.01,0.41298561096191405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.01,0.5505152225494385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,balanced,0.03263466556866964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.03856639862060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.039212799072265624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,balanced,0.04027733455101649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,balanced,0.07446399827798207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,balanced,0.08276799817879994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,balanced,0.10245866576830547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,balanced,0.12261866529782613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,balanced,0.16378666957219443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,balanced,0.2105813423792521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,balanced,0.2850933273633321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,balanced,0.37352001667022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.1602112054824829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.01,0.8972928047180175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.01,1.0767680168151856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.20127999782562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.2758847951889038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.08213120102882385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.11322879791259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.17377920150756837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.17290879487991334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,power_law_1.2,0.2662400007247925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,power_law_1.2,0.31781759262084963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,power_law_1.2,0.5436927795410156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.3766527891159058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,power_law_1.2,0.5828288078308106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.0711679995059967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.08972799777984619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.09928960204124451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.11603200435638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.13144960403442382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.17261439561843872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.21783039569854737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.2832063913345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.34012799263000487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.6021312236785888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.6206079959869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,balanced,0.04285866518815359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,balanced,0.07938133180141449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,balanced,0.11289599537849426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,balanced,0.11242133378982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,balanced,0.11340266466140747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,balanced,0.11515733599662781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,balanced,0.11229866743087769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,balanced,0.1149013340473175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,balanced,0.11706667145093282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,balanced,0.11658133069674174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,balanced,0.11902399857838948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,balanced,0.1255573332309723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,balanced,0.1267519990603129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,balanced,0.1302880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,balanced,0.13895466923713684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,balanced,0.1454026699066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,balanced,0.17659199237823486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,balanced,0.21158933639526367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,balanced,0.2376586596171061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,balanced,0.30856533845265705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,balanced,0.3736106554667155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,balanced,0.523423989613851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,balanced,0.6440853277842203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,balanced,0.09220799803733826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,balanced,0.1317866643269857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,balanced,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,balanced,0.16821332772572836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,balanced,0.17226133743921915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,balanced,0.17216000954310098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,balanced,0.17381866772969565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,balanced,0.18001067638397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,balanced,0.180351992448171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,balanced,0.18611733118693033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,balanced,0.19595734278361002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,balanced,0.2007360061009725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,balanced,0.21515733003616333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,balanced,0.24212799469629923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,balanced,0.26057066520055133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,balanced,0.3112799922625224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,balanced,0.35212798913319904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,balanced,0.44731732209523517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,balanced,0.5639946858088175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,balanced,0.749018669128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,balanced,0.9968266487121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,balanced,1.4253493944803874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,balanced,1.8287253379821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,balanced,0.05446400245030721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,balanced,0.0798773318529129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,balanced,0.16726932922999063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,balanced,0.1681279937426249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,balanced,0.16736533244450888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,balanced,0.17222400506337485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,balanced,0.17071467638015747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,balanced,0.17377066612243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,balanced,0.17783466974894205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,balanced,0.1809813380241394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,balanced,0.18499199549357095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,balanced,0.19477333625157675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,balanced,0.20218666394551596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,balanced,0.22316267093022665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,balanced,0.24589866399765015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,balanced,0.2629706660906474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,balanced,0.3063466747601827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,balanced,0.3500906626383464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,balanced,0.4843680063883464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,balanced,0.5446399847666422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,balanced,0.7738666534423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,balanced,0.9505759874979655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,balanced,1.3627413113911946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,balanced,1.759893258412679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,power_law_1.2,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,power_law_1.2,0.028300800919532777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,power_law_1.2,0.027065598964691163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,power_law_1.2,0.030015999078750612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,power_law_1.2,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,power_law_1.2,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,power_law_1.2,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,power_law_1.2,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,power_law_1.2,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,power_law_1.2,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,power_law_1.2,0.07988479733467102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,power_law_1.2,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,power_law_1.2,0.1434175968170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,power_law_1.2,0.2012864112854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,power_law_1.2,0.24666879177093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,power_law_1.2,0.33834240436553953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,balanced,0.05100266635417938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,balanced,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,balanced,0.069882666071256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,balanced,0.09726400176684062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,balanced,0.09806933005650838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,balanced,0.097461332877477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,balanced,0.09871466954549153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,balanced,0.09773332873980205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,balanced,0.09860266248385112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,balanced,0.10077866911888123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,balanced,0.10063466429710388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,balanced,0.09961066643397014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,balanced,0.10734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,balanced,0.10638399918874104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,balanced,0.10945600271224976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,balanced,0.11557333668073018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,balanced,0.11825066804885864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,balanced,0.12794666488965353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,balanced,0.13738666971524557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,balanced,0.1572533349196116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,balanced,0.17805333932240805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,balanced,0.21428799629211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,balanced,0.25358933210372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,balanced,0.3405226469039917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,balanced,0.42218132813771564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,power_law_1.2,0.4739520072937012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.01,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.01,0.05514240264892578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.01,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.01,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.01,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.01,0.06096000075340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.01,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.01,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.01,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.01,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.01,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.01,0.12655999660491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.01,0.1441856026649475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.01,0.18472319841384888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.01,0.18642560243606568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.01,0.2531840085983276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.01,0.2881407976150513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.01,0.39971199035644533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.01,0.4714303970336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.09649919867515563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.0961471974849701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.09970560073852539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.11368319988250733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.13370239734649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.14391679763793946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.1633344054222107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.18842240571975707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.05899519920349121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.1093567967414856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.13836159706115722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.17006080150604247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.23009281158447265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.2660799980163574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,power_law_1.01,0.3601919889450073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,power_law_1.01,0.4404416084289551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.23386878967285157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.27946879863739016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.3713023900985718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.4588223934173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.6414400100708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.809939193725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.1619135856628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,power_law_1.01,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,power_law_1.01,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.4876799583435059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,power_law_1.01,0.11398400068283081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,power_law_1.01,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,power_law_1.01,0.11740800142288207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,power_law_1.01,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,power_law_1.01,0.12314239740371705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,power_law_1.01,0.12388479709625244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,power_law_1.01,0.12815359830856324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,power_law_1.01,0.1324735999107361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,power_law_1.01,0.13950079679489136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,power_law_1.01,0.14826879501342774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,power_law_1.01,0.15216000080108644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,power_law_1.01,0.16216319799423218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,power_law_1.01,0.18005759716033937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,power_law_1.01,0.19719680547714233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,power_law_1.01,0.24531199932098388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,power_law_1.01,0.29009280204772947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,power_law_1.01,0.36129920482635497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,power_law_1.01,0.4491263866424561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,power_law_1.01,0.6216896057128907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,power_law_1.01,0.7842751979827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,power_law_1.01,1.1258815765380858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,power_law_1.01,1.3910207748413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,balanced,0.06798399984836578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,balanced,0.08592533071835835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,balanced,0.08786666393280029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,balanced,0.0869599978129069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,balanced,0.08855467041333516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,balanced,0.09119466940561931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,balanced,0.0997973382472992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,balanced,0.10005333026250203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,balanced,0.10435733199119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,balanced,0.11406399806340535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.12122666835784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.14082133769989014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.1560426652431488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.18917866547902426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.2197493314743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.2847200036048889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.3596479892730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.4875733455022176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.6228266557057699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.08014079928398132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.08239359855651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.09937279820442199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.1250048041343689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.13248000144958497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.1387712001800537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.16910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.20894079208374022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.24554879665374757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.3373248100280762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,power_law_1.01,0.43313279151916506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,power_law_1.01,0.5895103931427002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,power_law_1.01,0.7996223926544189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,balanced,0.026447998980681103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,balanced,0.027285332481066387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,balanced,0.0283146674434344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,balanced,0.027242665489514668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,power_law_1.01,1.0475711822509766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.04305280148983002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.10592000484466553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.14263039827346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,balanced,0.029994666576385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.1565951943397522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.20369279384613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.2501568078994751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.3664191961288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.4088575839996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.6425856113433838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,0.8337535858154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,power_law_1.2,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,power_law_1.2,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,power_law_1.2,0.11664639711380005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,power_law_1.2,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,power_law_1.2,0.11989760398864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,power_law_1.2,0.12579840421676636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,power_law_1.2,0.12732800245285034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,power_law_1.2,0.1319615960121155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,power_law_1.2,0.13579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,power_law_1.2,0.14146560430526733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,power_law_1.2,0.15029759407043458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,power_law_1.2,0.15967999696731566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,power_law_1.2,0.16420480012893676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,power_law_1.2,0.18269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,power_law_1.2,0.20726399421691893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,power_law_1.2,0.25806078910827634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,power_law_1.2,0.28089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,power_law_1.2,0.3616447925567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,power_law_1.2,0.4826176166534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,power_law_1.2,0.6274623870849609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,power_law_1.2,0.8255295753479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,power_law_1.01,0.061363202333450315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,power_law_1.01,0.07280640006065368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,power_law_1.01,0.10039680004119873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,power_law_1.01,0.10906239748001098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,power_law_1.01,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,power_law_1.01,0.13829120397567748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,power_law_1.01,0.13664640188217164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,power_law_1.01,0.13390079736709595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,power_law_1.01,0.13883520364761354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,power_law_1.01,0.14755200147628783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,power_law_1.01,0.14501760005950928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,power_law_1.01,0.15425920486450195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,power_law_1.01,0.16257920265197753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,power_law_1.01,0.16771199703216552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,power_law_1.01,0.17467520236968995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,power_law_1.01,0.19956480264663695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.01,0.22488958835601808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.01,0.26968960762023925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.01,0.3241215944290161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.01,0.41901440620422364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.01,0.5135680198669433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.01,0.7280320167541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,power_law_1.2,1.1686976432800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.01,0.947270393371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.01,1.319660758972168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.01,1.7272512435913085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,power_law_1.2,1.6059520721435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,balanced,0.08900266885757446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,balanced,0.10731200377146403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,balanced,0.14405866463979086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,balanced,0.1788960099220276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,balanced,0.0420959989229838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,balanced,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,balanced,0.09939199686050415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,balanced,0.10123733679453532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,balanced,0.10128000378608704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,balanced,0.10008000334103902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,balanced,0.10225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,balanced,0.10358400146166484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,balanced,0.10540266831715901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,balanced,0.11310399572054546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,balanced,0.11288533608118693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,balanced,0.11949333548545837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,balanced,0.12829867005348206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,balanced,0.13742933670679727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,balanced,0.1546293298403422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,balanced,0.17179199059804282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,balanced,0.2057759960492452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,balanced,0.2510133385658264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,balanced,0.320906658967336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,balanced,0.4070613384246826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,balanced,0.5655946731567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,balanced,0.7000053723653158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,power_law_1.01,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,power_law_1.01,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,power_law_1.01,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,power_law_1.01,0.06837760210037232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,power_law_1.01,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,power_law_1.01,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,power_law_1.01,0.137990403175354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,power_law_1.01,0.18471039533615113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,power_law_1.01,0.218886399269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,power_law_1.01,0.30106239318847655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,power_law_1.01,0.3896575927734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,balanced,0.09286933143933614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,balanced,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,balanced,0.11538133025169373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,balanced,0.11708266536394756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,balanced,0.11756799618403117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,balanced,0.11381866534550984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,balanced,0.11711999773979187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,balanced,0.11741866668065389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,balanced,0.12210133671760559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,balanced,0.12935466567675272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,balanced,0.13235732913017273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,balanced,0.14272532860438028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,balanced,0.15987733006477356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,balanced,0.1756640076637268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,balanced,0.20761599143346152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,balanced,0.23696533838907877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,balanced,0.3026026686032613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,balanced,0.3777279853820801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,balanced,0.5038400093714396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,balanced,0.6695893605550131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,balanced,0.947983980178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,balanced,1.2321759859720867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.0282943993806839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.029017600417137145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.0950976014137268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.13555840253829957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.15859199762344361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.236844801902771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,power_law_1.01,0.2856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,balanced,0.0352906659245491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,balanced,0.04773866633574168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,balanced,0.04744533201058706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,balanced,0.07069333394368489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,balanced,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,balanced,0.10244266192118327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,balanced,0.11717866857846577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,balanced,0.15218666195869446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,balanced,0.17914666732152304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,balanced,0.23861332734425864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,balanced,0.3031146725018819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.10110080242156982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.15796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.1831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.22478721141815186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.23006720542907716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.23912959098815917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.2388607978820801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.2500607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.26042239665985106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.26439039707183837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.272108793258667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.28487040996551516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.2936896085739136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.30908799171447754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.3570751905441284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.39480319023132326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.49008641242980955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5667520046234131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.7451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.9144703865051269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.2591232299804687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.607801628112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.3023807525634767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.0445632934570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,power_law_1.2,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,power_law_1.2,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,power_law_1.2,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,power_law_1.2,0.061740797758102414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,power_law_1.2,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,power_law_1.2,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,power_law_1.2,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,power_law_1.2,0.12535680532455445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,power_law_1.2,0.1559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,power_law_1.2,0.20273280143737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,power_law_1.2,0.2402048110961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,power_law_1.2,0.34549760818481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,power_law_1.2,0.4387455940246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,power_law_1.2,0.6607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,power_law_1.2,0.8643391609191895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,balanced,0.06454400221506755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,balanced,0.08299200236797333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,balanced,0.08638399839401245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,balanced,0.08563199639320374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,balanced,0.08889599641164143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,balanced,0.087909330924352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,balanced,0.08738666772842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,balanced,0.08913600444793701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,balanced,0.08870399991671245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,balanced,0.0892639954884847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,balanced,0.09674132863680522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,balanced,0.09739733735720317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,balanced,0.09900800387064616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,balanced,0.10000532865524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,balanced,0.09924266735712688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,balanced,0.11055999994277954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,balanced,0.1274720033009847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,balanced,0.13615999619166055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,balanced,0.18152532974878946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,balanced,0.21344000101089478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,balanced,0.2877066731452942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,balanced,0.3500639994939168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,balanced,0.5076160033543905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,balanced,0.6415786743164062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,balanced,0.9641653696695963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,balanced,1.2525066534678142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,power_law_1.2,0.02975359857082367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,power_law_1.2,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,power_law_1.2,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,power_law_1.2,0.07286400198936463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.2,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.2,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.2,0.14642560482025146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.2,0.22213120460510255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.2,0.2628159999847412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.2,0.411513614654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.2,0.5880767822265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.2,0.9926143646240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,power_law_1.01,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,power_law_1.01,0.02748799920082092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,power_law_1.01,0.027807998657226562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,power_law_1.01,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,power_law_1.01,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,power_law_1.01,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,power_law_1.01,0.13239680528640746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,power_law_1.01,0.17524479627609252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,power_law_1.01,0.23265280723571777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,power_law_1.01,0.298688006401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,power_law_1.01,0.5141759872436523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,power_law_1.2,0.06507520079612732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,power_law_1.2,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,power_law_1.2,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,power_law_1.2,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,power_law_1.2,0.06995199918746949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,power_law_1.2,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,power_law_1.2,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,power_law_1.2,0.09933440089225769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,power_law_1.2,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,power_law_1.2,0.13660160303115845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,power_law_1.2,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,power_law_1.2,0.2082047939300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,power_law_1.2,0.2549504041671753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.2,1.3434240341186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,power_law_1.2,0.3651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,power_law_1.2,0.464844799041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,power_law_1.2,0.6540287971496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,power_law_1.2,0.8858304023742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,power_law_1.01,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,power_law_1.01,0.029254400730133058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,power_law_1.01,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,power_law_1.01,0.03362559974193573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,power_law_1.01,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,power_law_1.01,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,power_law_1.01,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,power_law_1.01,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,power_law_1.01,0.1570688009262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,power_law_1.01,0.18529920578002929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,power_law_1.01,0.26156799793243407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,power_law_1.01,0.33704960346221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,power_law_1.01,0.49236478805541994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,power_law_1.01,0.641868782043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,balanced,0.029717333614826202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,balanced,0.030597334106763203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,balanced,0.03601066768169403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,balanced,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,balanced,0.09416533509890239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,balanced,0.1218293309211731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,balanced,0.14406933387120566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,balanced,0.19394665956497192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,balanced,0.2387359937032064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,power_law_1.2,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,power_law_1.2,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,power_law_1.2,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,power_law_1.2,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,power_law_1.2,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,power_law_1.2,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,power_law_1.2,0.08338559865951538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,power_law_1.2,0.0970624029636383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,power_law_1.2,0.11233919858932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,power_law_1.2,0.14589439630508422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,power_law_1.2,0.1751680016517639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,power_law_1.2,0.22613120079040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,power_law_1.2,0.2843071937561035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,power_law_1.2,0.38562560081481934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,power_law_1.2,0.5262080192565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,power_law_1.2,0.7457727909088134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,power_law_1.2,0.9541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.19671679735183717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.32522881031036377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.5872447967529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.020364761352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.01,1.4101311683654785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.01,2.364422416687012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.01,2.5350271224975587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.01,2.609958457946777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.01,2.683776092529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.01,2.7610496520996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.01,2.795091247558594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.01,2.901126480102539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.01,2.927686309814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.01,2.9863168716430666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.01,3.0725120544433593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.01,3.1631103515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.01,3.282495880126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.01,3.547711944580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.01,3.7021312713623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.01,4.051948928833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.01,4.084844970703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.01,4.710777664184571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.01,5.22790412902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.01,6.338784027099609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.01,8.488333129882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.01,10.972550201416016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.01,12.01122589111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,balanced,0.04744533201058706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,balanced,0.05093866586685181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,balanced,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,balanced,0.0563679983218511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,balanced,0.07072000205516815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,balanced,0.088837335507075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,balanced,0.0990826686223348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,balanced,0.1336373289426168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,balanced,0.1667626698811849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,balanced,0.23407467206319174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,balanced,0.2863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,balanced,0.4041386842727661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,balanced,0.5177066723505656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,balanced,0.07863999903202057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,balanced,0.08042133351167043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,balanced,0.07938133180141449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,balanced,0.08121066788832347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,balanced,0.08082133531570435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,balanced,0.08255999783674876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,balanced,0.08429867029190063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,balanced,0.08468799789746602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,balanced,0.09361066420873006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,balanced,0.09586133559544881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,balanced,0.10507733623186748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,balanced,0.11221333344777425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,balanced,0.12330133716265361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,balanced,0.13784533739089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,balanced,0.15364266435305277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,balanced,0.1943733294804891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,balanced,0.22407466173171997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,balanced,0.2977173328399658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,balanced,0.38888001441955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,balanced,0.5454293489456177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,balanced,0.7141760190327963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,power_law_1.2,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,power_law_1.2,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,power_law_1.2,0.03681919872760773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,power_law_1.2,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,power_law_1.2,0.07653759717941284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,power_law_1.2,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,power_law_1.2,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,power_law_1.2,0.16839040517807008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,power_law_1.2,0.19829119443893434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,power_law_1.2,0.2708031892776489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,power_law_1.2,0.3889087915420532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,power_law_1.2,0.5169856071472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.14076800346374513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.1416640043258667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.13893760442733766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.12391680479049683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.12591359615325928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.13283200263977052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.13824000358581542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.1370303988456726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.13872640132904052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.1353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.15802240371704102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.15530240535736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.1667199969291687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.16530560255050658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.17600640058517455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.18017280101776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.20743680000305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.2169663906097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.27302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.30570240020751954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.376313591003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,power_law_1.01,0.4592319965362549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,power_law_1.01,0.6209792137145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,power_law_1.01,0.7910079956054688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,balanced,0.05026133358478546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,balanced,0.052442664901415505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,balanced,0.059792002042134605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,balanced,0.06588799754778545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,balanced,0.06819733480612437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,balanced,0.0746613343556722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.07834666470686595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.09547199805577596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.11010133226712544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.1335093379020691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.1549493372440338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.1995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.24523200591405234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.3357386589050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.4182986815770467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,power_law_1.2,0.6122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,balanced,0.04027733455101649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,balanced,0.06052800019582113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,balanced,0.06660800178845723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,balanced,0.06449066599210103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,balanced,0.06472533444563548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,balanced,0.06595733265082042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,balanced,0.07791466514269511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,balanced,0.0825386643409729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,balanced,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,balanced,0.11274133125940959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,balanced,0.12616533041000366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,balanced,0.16318399707476297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,balanced,0.19384000698725382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,balanced,0.2627253333727519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,balanced,0.3218453327814738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,balanced,0.4542880058288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,balanced,0.5743573506673177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,power_law_1.01,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,power_law_1.01,0.10769280195236205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,power_law_1.01,0.12150399684906006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,power_law_1.01,0.14543999433517457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,power_law_1.01,0.1481279969215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,power_law_1.01,0.14956799745559693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,power_law_1.01,0.15943039655685426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,power_law_1.01,0.16069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,power_law_1.01,0.16666239500045776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,power_law_1.01,0.16673280000686647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,power_law_1.01,0.17354240417480468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,power_law_1.01,0.18069759607315064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,power_law_1.01,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,power_law_1.01,0.19219839572906494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,power_law_1.01,0.2111232042312622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,power_law_1.01,0.2265536069869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,power_law_1.01,0.2649343967437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,power_law_1.01,0.28752000331878663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,power_law_1.01,0.3589632034301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,power_law_1.01,0.412332820892334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,power_law_1.01,0.5321599960327148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,power_law_1.01,0.6617856025695801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,power_law_1.01,0.8939264297485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,power_law_1.01,1.1302080154418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,power_law_1.01,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,power_law_1.01,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,power_law_1.01,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,power_law_1.01,0.08611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,power_law_1.01,0.08725759983062745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,power_law_1.01,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,power_law_1.01,0.09925119876861573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,power_law_1.01,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,power_law_1.01,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,power_law_1.01,0.1184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,power_law_1.01,0.14142080545425414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,power_law_1.01,0.15688960552215575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,power_law_1.01,0.17953280210494996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,power_law_1.01,0.2117759943008423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,power_law_1.01,0.26860799789428713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.09934719800949096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,power_law_1.01,0.3371392011642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.10476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,power_law_1.01,0.4484416007995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,power_law_1.01,0.5645503997802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.13574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,power_law_1.01,0.8354559898376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.17380479574203492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,power_law_1.01,1.0357824325561524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.20734720230102538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.22897920608520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.2672384023666382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.2789695978164673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.2821439981460571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.2893183946609497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.30536320209503176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.3115135908126831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.3234240055084229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.3342911958694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3446592092514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.3638783931732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.41665282249450686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.4493055820465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,0.5454527854919433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,0.6299903869628907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,0.8193663597106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,0.9955007553100585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,1.366220760345459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.01,1.7288320541381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.01,2.456390380859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.01,3.189606475830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,balanced,0.09822400410970052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,balanced,0.09713066617647807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,balanced,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,balanced,0.0990826686223348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,balanced,0.09873066345850627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,balanced,0.09999466935793559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,balanced,0.10058666268984477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,balanced,0.10175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,balanced,0.10211199522018433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,balanced,0.10436800122261047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,balanced,0.10610666871070862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,balanced,0.11343466242154439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,balanced,0.11580800016721089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,balanced,0.11742933591206868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,balanced,0.14777599771817526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,balanced,0.15738667050997415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,balanced,0.2021920084953308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,balanced,0.24642133712768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,balanced,0.3416106700897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,balanced,0.45685335000356037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,balanced,0.6526026725769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,balanced,0.8537973562876383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,balanced,1.231061299641927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,balanced,1.6079893112182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.049907198548316954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.05248640179634094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.06446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.1850559949874878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.24906880855560304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.3107583999633789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.44538240432739257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.058815997838973996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.06620799899101257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.5547840118408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.09900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.10712959766387939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.1134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.12545280456542968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.1298624038696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.135808002948761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.17200000286102296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.20469119548797607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.23432960510253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.29900801181793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.381823992729187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.5095551967620849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,power_law_1.01,0.5899392127990722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,power_law_1.01,0.8774335861206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,power_law_1.01,1.1186880111694335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,balanced,0.03179733455181122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,balanced,0.039061332742373146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,balanced,0.058575997749964394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,balanced,0.06579199930032094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,balanced,0.0823466678460439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,balanced,0.12231466174125671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,balanced,0.14630400141080221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,balanced,0.19926400979359946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,balanced,0.24559466044108072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,power_law_1.01,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,power_law_1.01,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,power_law_1.01,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,power_law_1.01,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,power_law_1.01,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,power_law_1.01,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,power_law_1.01,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,power_law_1.01,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,power_law_1.01,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,power_law_1.01,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,power_law_1.01,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.01,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.01,0.11573120355606079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.01,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.01,0.16757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.01,0.1964416027069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.01,0.25813119411468505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.01,0.3409663915634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.01,0.48326401710510253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.01,0.5707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.09520000219345093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.09868800044059753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.1046336054801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.10730240345001221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.13907840251922607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.15518079996109008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.17640960216522217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.18623360395431518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.2382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.28834559917449953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.3788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5032063961029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.7139647960662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.819923210144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.1834752082824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.4826623916625976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,balanced,0.04808000226815542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,balanced,0.05043200155099233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,balanced,0.0557226687669754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,balanced,0.059893334905306496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,balanced,0.11425600449244182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,balanced,0.14149333039919534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,balanced,0.14289599657058716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,balanced,0.14563733339309692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,balanced,0.14562132954597473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,balanced,0.14815466602643332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,balanced,0.15401066342989603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,balanced,0.15470932920773825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,balanced,0.1585546632607778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,balanced,0.16773333152135214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,balanced,0.1728853384653727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,balanced,0.18314133087793985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,balanced,0.21359467506408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,balanced,0.23287999629974365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,balanced,0.2804693380991618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,balanced,0.3227573235829671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,balanced,0.4115999937057495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,balanced,0.528714656829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,balanced,0.7101600170135498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,balanced,0.9642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,balanced,1.3815466562906902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,balanced,1.7852320671081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,power_law_1.01,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,power_law_1.01,0.106278395652771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,power_law_1.01,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,power_law_1.01,0.1241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,power_law_1.01,0.12533119916915894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,power_law_1.01,0.12666879892349242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,power_law_1.01,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,power_law_1.01,0.13322880268096923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,power_law_1.01,0.14380160570144654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,power_law_1.01,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,power_law_1.01,0.14783999919891358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,power_law_1.01,0.15251840353012086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,power_law_1.01,0.1583295941352844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,power_law_1.01,0.1698815941810608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,power_law_1.01,0.1825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,power_law_1.01,0.18819199800491332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,power_law_1.01,0.2160383939743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,power_law_1.01,0.2620032072067261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,power_law_1.01,0.3158015966415405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,power_law_1.01,0.4127744197845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,power_law_1.01,0.4632512092590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,power_law_1.01,0.6518784046173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,power_law_1.01,0.9052543640136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,power_law_1.01,1.1889216423034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,power_law_1.01,1.634022331237793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,power_law_1.01,2.7631296157836913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,power_law_1.01,3.2998462677001954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.07511680126190186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.08195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.13936640024185182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.16090879440307618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.2186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.2637824058532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.3452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.4683839797973633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.6332223892211915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.9350208282470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.12256640195846558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.11584000587463379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.11749119758605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.12434560060501099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.11788159608840942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.12716799974441528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.12729599475860595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.14052480459213257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.16311039924621581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.18534400463104247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.2170111894607544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.22144639492034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.2649791955947876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.3065664052963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.4177152156829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,power_law_1.01,0.475219202041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,power_law_1.01,0.7360767841339111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,power_law_1.01,0.8072575569152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,power_law_1.01,0.10359040498733521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,power_law_1.01,0.17289600372314454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,power_law_1.01,0.2615488052368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,power_law_1.01,0.27752320766448973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,power_law_1.01,0.28665599822998045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,power_law_1.01,0.29035520553588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,power_law_1.01,0.29704320430755615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,power_law_1.01,0.30420479774475095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,power_law_1.01,0.3154304027557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,power_law_1.01,0.32467200756073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,power_law_1.01,0.335097599029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,power_law_1.01,0.3455359935760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,power_law_1.01,0.36625280380249026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,power_law_1.01,0.38029439449310304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,power_law_1.01,0.42169599533081054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,power_law_1.01,0.4755072116851807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,power_law_1.01,0.5865024089813232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,power_law_1.01,0.7085184097290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,power_law_1.01,0.9003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,power_law_1.01,1.0161919593811035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,power_law_1.01,1.3425984382629395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,power_law_1.01,1.742233657836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,power_law_1.01,2.391007995605469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,power_law_1.01,3.079974365234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.029286399483680725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.049830400943756105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.08995839953422546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.12594560384750367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.17566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,power_law_1.01,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,power_law_1.01,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,power_law_1.01,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,power_law_1.01,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,power_law_1.01,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,power_law_1.01,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,power_law_1.01,0.05123199820518494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,power_law_1.01,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,power_law_1.01,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,power_law_1.01,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.01,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.01,0.10543999671936036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.01,0.13623679876327516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.01,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.01,0.2232448101043701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.01,0.2709376096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.01,0.38525440692901614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.01,0.5332032203674316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.29596800804138185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,power_law_1.2,0.4838079929351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,balanced,0.07060266534487407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,balanced,0.09743466973304749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,balanced,0.14654399951299033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,balanced,0.24347732464472452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,balanced,0.4386719862620036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,balanced,0.6342346668243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,balanced,0.6350666681925455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,balanced,0.6367573340733846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,balanced,0.636522650718689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,balanced,0.6393493413925171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,balanced,0.6425706545511881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,balanced,0.645306666692098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,balanced,0.6466240088144938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,balanced,0.6571893294652303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,balanced,0.6607626676559448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,balanced,0.6682720184326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,balanced,0.6832373142242432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,balanced,0.6956160068511963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,balanced,0.7264266808827718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,balanced,0.7561972935994467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,balanced,0.8103840351104736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,balanced,0.8705493609110514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,balanced,0.9710720380147299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,balanced,1.125381310780843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,balanced,1.3233493169148762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,balanced,1.7176532745361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,balanced,0.0555626650651296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,balanced,0.07161066432793935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,balanced,0.09334933757781982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,balanced,0.1337279975414276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,balanced,0.17063466707865396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,balanced,0.24753065903981528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,balanced,0.32924799124399823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,balanced,0.4872693220774333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,balanced,0.6318133274714152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,power_law_1.2,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,power_law_1.2,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,power_law_1.2,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,power_law_1.2,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,power_law_1.2,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,power_law_1.2,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.2,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.2,0.11942399740219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.2,0.16509439945220947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.2,0.23552000522613525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.2,0.29838080406188966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.2,0.4428351879119873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.2,0.5548927783966064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.09383040070533752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.09244160056114196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.09572479724884034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.10008319616317748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.10039680004119873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.09631360173225403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.0991487979888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.15397119522094727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.1602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.18243199586868286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.24675838947296141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.30409600734710696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,power_law_1.2,0.4633344173431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,power_law_1.2,0.5837696075439454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,power_law_1.2,0.8824704170227051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,power_law_1.2,1.1530752182006836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.2,0.9779328346252442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.2,1.3085503578186035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,power_law_1.01,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,power_law_1.01,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,power_law_1.01,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,power_law_1.01,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,power_law_1.01,0.09224320054054261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,power_law_1.01,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,power_law_1.01,0.09591040015220642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,power_law_1.01,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,power_law_1.01,0.10503679513931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,power_law_1.01,0.11527040004730224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,power_law_1.01,0.11783679723739623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,power_law_1.01,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,power_law_1.01,0.14333440065383912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,power_law_1.01,0.15845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,power_law_1.01,0.19223040342330933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,power_law_1.01,0.22261760234832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,power_law_1.01,0.29957120418548583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,power_law_1.01,0.35697920322418214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,power_law_1.01,0.4981823921203613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,power_law_1.01,0.6270847797393799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,power_law_1.01,0.9131903648376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,power_law_1.01,1.1772800445556642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,balanced,0.05442133545875549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,balanced,0.08311999837557475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,balanced,0.1258080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,balanced,0.18479466438293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,balanced,0.213919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,balanced,0.21529599030812582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,balanced,0.21842666467030844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,balanced,0.21971199909845987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,balanced,0.21847466627756754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,balanced,0.22325332959493002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,balanced,0.22444266080856323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,balanced,0.22427733739217123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,balanced,0.22904000679651895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,balanced,0.23449599742889404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,balanced,0.23761065800984701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,balanced,0.2464266618092855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,balanced,0.257968008518219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.2689653237660726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.32976533969243366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.3242453336715698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.43108801047007245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.4429119825363159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.6370293299357096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,balanced,0.735706647237142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.040885368982951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,balanced,1.3066666920979817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,power_law_1.2,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,power_law_1.2,0.10635520219802856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,power_law_1.2,0.15141760110855101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,power_law_1.2,0.20606720447540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,power_law_1.2,0.2659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,power_law_1.2,0.4523712158203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,power_law_1.2,0.4788991928100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,power_law_1.2,0.5039872169494629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,power_law_1.2,0.5322432041168212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,power_law_1.2,0.541644811630249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,power_law_1.2,0.5408512115478515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,power_law_1.2,0.5719359874725342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,power_law_1.2,0.5753471851348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,power_law_1.2,0.6050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,power_law_1.2,0.6174848079681396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,power_law_1.2,0.643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,power_law_1.2,0.6638720035552979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,power_law_1.2,0.7446847915649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,power_law_1.2,0.7802815914154053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,power_law_1.2,0.920531177520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,power_law_1.2,1.0091903686523438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,power_law_1.2,1.2699007987976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,power_law_1.2,1.3847999572753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,power_law_1.2,1.797804832458496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,power_law_1.2,2.4945920944213866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,power_law_1.2,3.2827392578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,power_law_1.2,4.034969711303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.01,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.01,0.07737600207328796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.01,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.01,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.01,0.07740799784660339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.01,0.08073599934577942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.01,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.01,0.08796799778938294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.01,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.01,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.01,0.1286784052848816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.01,0.16560640335083007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.01,0.18989440202713012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.01,0.2430016040802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.01,0.26194560527801514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.01,0.3524672031402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.01,0.42639999389648436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.01,0.5723264217376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.01,0.7053696155548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.01,0.1895616054534912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.01,0.31424000263214114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.01,0.32242560386657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.01,0.4463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.01,0.5564032077789307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.01,0.7121344089508057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.01,0.7291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.01,0.6963071823120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.01,0.7373760223388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.01,0.7390655994415283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.01,0.7501503944396972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.01,0.7621439933776856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.01,0.7995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.01,0.8098943710327149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.01,0.8165504455566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.01,0.8366975784301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.01,0.8857791900634766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.01,0.9919808387756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.01,1.0017536163330079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.01,1.1081088066101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.01,1.2137727737426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.01,1.4303296089172364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.01,1.515283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.01,1.8595327377319335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.01,2.4572223663330077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.01,3.125132751464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.01,3.672800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,power_law_1.01,0.023737600445747374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,power_law_1.01,0.024051199853420257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,power_law_1.01,0.02412160038948059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,power_law_1.01,0.02802560031414032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,power_law_1.01,0.024486400187015533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,power_law_1.01,0.025011199712753295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,power_law_1.01,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,power_law_1.01,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.10037120580673217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.10314879417419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.14661760330200196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.2132352113723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,balanced,0.04906133313973745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,balanced,0.07425599793593089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,balanced,0.08636800448099773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,balanced,0.08665600419044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,balanced,0.08707200487454732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,balanced,0.08654399712880452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,balanced,0.08887466788291931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,balanced,0.09063999851544698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,balanced,0.08938133716583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,balanced,0.0935093363126119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,balanced,0.09748799602190654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,balanced,0.0997226635615031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,balanced,0.10526933272679646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,balanced,0.11532800396283467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,balanced,0.12499200304349263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,balanced,0.1511146624883016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,balanced,0.17063466707865396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,balanced,0.20916799704233804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,balanced,0.2580159902572632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,balanced,0.3343626658121745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,balanced,0.43861865997314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,balanced,0.6015786727269491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,balanced,0.7858346303304037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,balanced,0.031136001149813335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,balanced,0.030661332110563915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,balanced,0.03305600086847941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,balanced,0.03826666623353958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,balanced,0.04953599969546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,balanced,0.05031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,balanced,0.05374933282534281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,balanced,0.06610133250554402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,balanced,0.09063466389973958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,balanced,0.11525332927703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,balanced,0.14010666807492575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,balanced,0.18919465939203897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,balanced,0.2323413292566935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,balanced,0.06046933432420095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,balanced,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,balanced,0.07239466905593872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,balanced,0.07869866490364075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,balanced,0.07295466462771098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,balanced,0.0729066679875056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,balanced,0.07681599756081899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,balanced,0.07970666885375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,balanced,0.08292800188064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,balanced,0.08470400174458821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,balanced,0.09070932865142822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,balanced,0.09423466523488362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,balanced,0.09934932986895244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,balanced,0.10537599523862202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,balanced,0.12328533331553142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,balanced,0.14013866583506265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,balanced,0.17428799470265707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,balanced,0.20735466480255127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,balanced,0.27609066168467206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,balanced,0.33904000123341876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,balanced,0.48044268290201825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,balanced,0.6091733376185099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,power_law_1.2,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,power_law_1.2,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,power_law_1.2,0.08352000117301941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,power_law_1.2,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,power_law_1.2,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,power_law_1.2,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,power_law_1.2,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,power_law_1.2,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,power_law_1.2,0.11614079475402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,power_law_1.2,0.12001279592514039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,power_law_1.2,0.12450560331344604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,power_law_1.2,0.1444159984588623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,power_law_1.2,0.16272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,power_law_1.2,0.19326080083847047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,power_law_1.2,0.2277695894241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,power_law_1.2,0.29991679191589354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,power_law_1.2,0.37411201000213623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,power_law_1.2,0.5095232009887696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,power_law_1.2,0.6665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,power_law_1.2,0.9441216468811036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,power_law_1.2,1.2599743843078612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,power_law_1.2,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,power_law_1.2,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,power_law_1.2,0.0377344012260437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,power_law_1.2,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,power_law_1.2,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,power_law_1.2,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,power_law_1.2,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,power_law_1.2,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,power_law_1.2,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.12960000038146974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.14028160572052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,power_law_1.2,0.09279360175132752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,power_law_1.2,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,power_law_1.2,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.1800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,power_law_1.2,0.09510400295257568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,power_law_1.2,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,power_law_1.2,0.0957759976387024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,power_law_1.2,0.09739519953727722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,power_law_1.2,0.09989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,power_law_1.2,0.10130560398101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,power_law_1.2,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,power_law_1.2,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,power_law_1.2,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,power_law_1.2,0.12124160528182984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,power_law_1.2,0.13415679931640626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.2,0.1379968047142029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.2,0.1795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.2,0.20030720233917237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.2,0.25345919132232664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.2,0.32892799377441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.2,0.44518399238586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.2,0.5941440105438233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.2,0.8574655532836915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.2,1.2647551536560058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.23658881187438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,power_law_1.2,0.2991679906845093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,power_law_1.2,0.4366591930389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,balanced,0.029557332396507263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,balanced,0.03172266731659571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,balanced,0.04227200150489807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,balanced,0.06503466765085857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,balanced,0.0687253326177597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,balanced,0.08828266461690266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,balanced,0.10127466917037964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,balanced,0.12590932846069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,balanced,0.15364799896876016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,balanced,0.20619734128316244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,balanced,0.2560960054397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,balanced,0.36609601974487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,balanced,0.4647573232650757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,balanced,0.08291733264923096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,balanced,0.1206826666990916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,balanced,0.1883253256479899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,balanced,0.3293973406155904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,balanced,0.5952426592508951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,balanced,0.8504266738891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,balanced,0.8575146993001302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,balanced,0.861962636311849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,balanced,0.866159995396932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,balanced,0.8674773375193278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,balanced,0.8751680056254069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,balanced,0.8845706780751547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,balanced,0.8903626600901285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,balanced,0.8987680276234945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,balanced,0.9146239757537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,balanced,0.923845370610555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,balanced,0.9497173627217611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,balanced,1.0046506722768147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,balanced,1.0569067001342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,balanced,1.3643520673116047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,balanced,1.2521973450978596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,balanced,1.820309321085612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,balanced,1.7199734052022297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,balanced,2.5381333033243814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,balanced,2.8089812596639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,balanced,4.338853200276692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,balanced,4.948090553283691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,balanced,0.0351946676770846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,balanced,0.03444266567627589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,balanced,0.03422933320204417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,balanced,0.03508266558249792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,balanced,0.04065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,balanced,0.044112001856168113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,balanced,0.050255998969078064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,balanced,0.06806399921576183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,balanced,0.0745600014925003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,balanced,0.10079999764760335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,balanced,0.11623467008272807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,balanced,0.1547040045261383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,balanced,0.18387732903162637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.10308480262756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.12327680587768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.1259775996208191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.13682559728622437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.1631872057914734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.18890880346298217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.23777918815612792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.28752000331878663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.39518721103668214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.49840641021728516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.7112192153930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.9167360305786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.327244758605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.7368959426879882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,balanced,0.07450133562088013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,balanced,0.13242133458455405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,balanced,0.19356266657511392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,balanced,0.31817599137624103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,balanced,0.44598933060963947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,balanced,0.44655466079711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,balanced,0.4493759870529175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,balanced,0.45209066073099774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,balanced,0.45318400859832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,balanced,0.46034133434295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,balanced,0.4621066649754842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,balanced,0.4678560098012288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,balanced,0.4788373311360677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,balanced,0.4811626672744751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,balanced,0.5047039985656738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,balanced,0.5229440132776896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.5584266583124796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.6109066804250082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.6494933366775513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.7480853398640951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,balanced,0.8480426470438639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,balanced,1.0331839720408122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,balanced,1.3231360117594402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,balanced,1.6986239751180012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,balanced,2.3164213498433432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,power_law_1.01,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,power_law_1.01,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,power_law_1.01,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,power_law_1.01,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,power_law_1.01,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,power_law_1.01,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,power_law_1.01,0.07036799788475037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,power_law_1.01,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,power_law_1.01,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,power_law_1.01,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,power_law_1.01,0.1272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,power_law_1.01,0.15533440113067626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,power_law_1.01,0.23066880702972412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,power_law_1.01,0.2825216054916382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,power_law_1.01,0.4018432140350342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,power_law_1.01,0.518336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.08451200127601624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.08926720023155213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.0953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.10485119819641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.10994559526443481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.11885440349578857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.12963839769363403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.14325120449066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.17754240036010743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.21913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.27197439670562745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.32843520641326907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.44206719398498534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,power_law_1.01,0.5364352226257324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,power_law_1.01,0.7591360092163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,power_law_1.01,1.106175994873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,balanced,0.07738666733105977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,balanced,0.12139200170834859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,balanced,0.20386133591334024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,balanced,0.37111465136210126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,balanced,0.37493332227071124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,balanced,0.38044798374176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,balanced,0.3781919876734416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,balanced,0.3815573453903198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,balanced,0.3832639853159587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,balanced,0.3878399928410848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,balanced,0.3918079932530721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,balanced,0.39695998032887775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,balanced,0.40199466546376544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,balanced,0.4111520051956177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,balanced,0.4194399913152059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,balanced,0.43355198701222736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,balanced,0.47129066785176593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,balanced,0.5080906550089518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,balanced,0.568618655204773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,balanced,0.6124533414840698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,balanced,0.7789813677469889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,balanced,0.8741013209025065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,balanced,1.2251946926116943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,balanced,1.4641812642415364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,balanced,2.090709368387858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,balanced,2.7024319966634116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,balanced,0.07832000156243642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,balanced,0.08507733543713887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,balanced,0.08602666854858398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,balanced,0.08564266562461853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,balanced,0.09014399846394856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,balanced,0.09223467111587524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,balanced,0.09790933132171631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,balanced,0.0995360016822815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,balanced,0.10293866197268169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,balanced,0.11732799808184306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,balanced,0.12634133299191794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,balanced,0.15685866276423135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,balanced,0.1790613333384196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,balanced,0.24003199736277261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,balanced,0.2340373396873474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,balanced,0.3142506678899129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,balanced,0.37645332018534344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,balanced,0.5290613174438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,balanced,0.6426986853281657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.0729856014251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.08014079928398132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.08019840121269226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.07813760042190551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.08071039915084839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.1083456039428711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.14042880535125732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.1411712050437927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.17840640544891356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.21367039680480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.30135679244995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,power_law_1.01,0.3831423997879028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,power_law_1.01,0.5382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,power_law_1.01,0.7089407920837403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,balanced,0.05017066498597463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,balanced,0.06257600088914235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,balanced,0.08089066545168559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,balanced,0.08103999992211659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,balanced,0.08255999783674876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,balanced,0.08425066868464152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,balanced,0.0846026639143626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,balanced,0.08674133817354839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,balanced,0.08844266335169475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,balanced,0.08902933200200398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,balanced,0.09418666362762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,balanced,0.09780800342559814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,balanced,0.1016373336315155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,balanced,0.10991467038790385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,balanced,0.12482666969299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,balanced,0.1357439955075582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,balanced,0.162800004084905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,balanced,0.19359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,balanced,0.24555732806523642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,balanced,0.29996800422668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,balanced,0.4408053159713745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,balanced,0.5428053140640259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,balanced,0.8078773021697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,balanced,1.0543306668599446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,balanced,0.061941335598627724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,balanced,0.06366933385531108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,balanced,0.06379733482996623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,balanced,0.06421866516272227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,balanced,0.07055466870466869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,balanced,0.07223466535409291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,balanced,0.07892799874146779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,balanced,0.08885866403579712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,balanced,0.09341866771380107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,balanced,0.10742933551470439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,balanced,0.12001599868138631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,balanced,0.15338666240374246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,balanced,0.17612266540527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,balanced,0.23948800563812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,balanced,0.28844799598058063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.12840319871902467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.15122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.22750720977783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.2304447889328003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.24382081031799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.2457535982131958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.24864640235900878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.24992640018463136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.25233919620513917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.25686399936676024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.26565120220184324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.27075839042663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.2797696113586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.30368640422821047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.30709760189056395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.32727680206298826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.394649600982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.44778881072998045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.5601344108581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.6699135780334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,0.8855999946594239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,1.0988096237182616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,1.5610624313354493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,1.989311981201172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,power_law_1.01,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,power_law_1.01,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,power_law_1.01,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,power_law_1.01,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,power_law_1.01,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,power_law_1.01,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,power_law_1.01,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,power_law_1.01,0.07237759828567505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,power_law_1.01,0.07820159792900086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,power_law_1.01,0.09896320104598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,power_law_1.01,0.11512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,power_law_1.01,0.13182719945907592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,power_law_1.01,0.16581120491027831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,power_law_1.01,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,power_law_1.01,0.20814080238342286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,power_law_1.01,0.2952192068099976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,power_law_1.01,0.348799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,power_law_1.01,0.49567999839782717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,power_law_1.01,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,power_law_1.01,0.644985580444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,power_law_1.01,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,2.870182418823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,power_law_1.01,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,power_law_1.01,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,power_law_1.01,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,power_law_1.01,0.06099839806556702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.01,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.01,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.01,0.14944640398025513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.01,0.18518400192260742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.01,0.24077439308166504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.01,0.29637119770050047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.01,0.44946560859680174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.01,0.6034111976623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,3.783251190185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.08903040289878845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.11029119491577148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.14244480133056642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.19524480104446412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.24041600227355958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.27171199321746825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.3198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,balanced,0.04502933224042257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.3317823886871338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,balanced,0.07357333103815715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,balanced,0.1056106686592102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,balanced,0.10682666301727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,balanced,0.10620799660682678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.3371903896331787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,balanced,0.10644267002741496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,balanced,0.10805333654085796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,balanced,0.10838933785756429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,balanced,0.11546132961908977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.347871994972229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,balanced,0.11423466602961223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,balanced,0.11806399623552959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,balanced,0.12959999839464822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,balanced,0.1344266633192698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,balanced,0.15204266707102457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.3639872074127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,balanced,0.16953066984812418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,balanced,0.20266133546829224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,balanced,0.23525333404541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.3720383882522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,balanced,0.2974986632664998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,balanced,0.36690131823221844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.3863552093505859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,balanced,0.5073759953180949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,balanced,0.6296639839808146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.3985215902328491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.4080959796905518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.43057918548583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.47728638648986815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.5166336059570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,0.6139264106750488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,0.6980480194091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,0.8769023895263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,1.0821056365966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,1.4528256416320802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.01,1.8647872924804687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.01,2.565088081359863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.01,3.2716670989990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,power_law_1.2,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,power_law_1.2,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,power_law_1.2,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,power_law_1.2,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,power_law_1.2,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,power_law_1.2,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,power_law_1.2,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,power_law_1.2,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,power_law_1.2,0.11535999774932862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,power_law_1.2,0.1386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,power_law_1.2,0.1834496021270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,power_law_1.2,0.25421440601348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,power_law_1.2,0.33644800186157225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,power_law_1.2,0.4050943851470947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,power_law_1.2,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,power_law_1.2,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,power_law_1.2,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,power_law_1.2,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,power_law_1.2,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,power_law_1.2,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,power_law_1.2,0.10742399692535401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,power_law_1.2,0.10999679565429688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,power_law_1.2,0.11738239526748658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,power_law_1.2,0.13586560487747193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,power_law_1.2,0.15708800554275512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,power_law_1.2,0.19475200176239013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,power_law_1.2,0.22442240715026857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,power_law_1.2,0.2974208116531372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,power_law_1.2,0.3780992031097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,power_law_1.2,0.5427135944366455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,power_law_1.2,0.706822395324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,power_law_1.2,1.013862419128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,power_law_1.2,1.2845952033996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,power_law_1.2,0.6314239978790284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,power_law_1.2,0.8796031951904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,power_law_1.01,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,power_law_1.01,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,power_law_1.01,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,power_law_1.01,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,power_law_1.01,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,power_law_1.01,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,power_law_1.01,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,power_law_1.01,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,power_law_1.01,0.09013760089874268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,power_law_1.01,0.11450239419937133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,power_law_1.01,0.13955199718475342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,power_law_1.01,0.19915519952774047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,power_law_1.01,0.25218560695648196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,power_law_1.01,0.34917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,power_law_1.01,0.45725440979003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.11898880004882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.14125440120697022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.15096960067749024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.18064639568328858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.18084479570388795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.18392959833145142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.18314239978790284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.19264639616012574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.1916864037513733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.20245759487152098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.21262080669403077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.21773440837860109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.23677439689636232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.2587199926376343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.2864831924438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.34549119472503664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.4069119930267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.520249605178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.6493375778198243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,0.8729727745056153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.0584511756896973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,1.5006400108337403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,1.9830848693847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.01,0.11534080505371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.01,0.14368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.01,0.13864320516586304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.01,0.15148160457611085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.01,0.13930879831314086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.01,0.14350080490112305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.01,0.13810559511184692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.01,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.01,0.14574079513549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.01,0.15271040201187133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.01,0.14839680194854737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.01,0.15530879497528077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.01,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.01,0.16587519645690918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.01,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.01,0.20090880393981933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.01,0.2126784086227417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.01,0.24703359603881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.01,0.2893824100494385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.01,0.3657792091369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.01,0.35878400802612304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.01,0.475219202041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.01,0.5793920040130616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.01,0.7980671882629394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.01,0.9716032028198243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,power_law_1.2,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,power_law_1.2,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,power_law_1.2,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,power_law_1.2,0.047603198885917665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,power_law_1.2,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,power_law_1.2,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,power_law_1.2,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,power_law_1.2,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.2,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.2,0.11452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.2,0.16476800441741943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.2,0.18129279613494872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.2,0.25333120822906496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.2,0.346777606010437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.2,0.5291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.2,0.7479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,power_law_1.2,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,power_law_1.2,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,power_law_1.2,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,power_law_1.2,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,power_law_1.2,0.09056000113487243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,power_law_1.2,0.09462400078773499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,power_law_1.2,0.12675199508666993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,power_law_1.2,0.14416639804840087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,power_law_1.2,0.18448640108108522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,power_law_1.2,0.21766400337219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,power_law_1.2,0.30081920623779296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,power_law_1.2,0.3841471910476685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,power_law_1.2,0.5478464126586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,power_law_1.2,0.7119167804718017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,balanced,0.03798400113979975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,balanced,0.06569066643714905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,balanced,0.06712000072002411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,balanced,0.06790933509667714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,balanced,0.06701333324114482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,balanced,0.0684799998998642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,balanced,0.06902400155862172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,balanced,0.07277866701285045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,balanced,0.0820853312810262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,balanced,0.08264000217119853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,balanced,0.09839466214179993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.11622400085131328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.13285332918167114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.14641066392262778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.18170666694641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.22157333294550577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.29266132911046344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.36640532811482746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.5200373331705729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.6695146560668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,power_law_1.2,1.0472000122070313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,balanced,0.04256533086299896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,balanced,0.042021334171295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,balanced,0.044069334864616394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,balanced,0.045552000403404236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,balanced,0.046581332882245384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06063466767470042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,balanced,0.06843199829260509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.0803413341442744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.10000532865524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.12076266606648763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.14428800344467163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.18868800004323324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.23276267449061075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.32421332597732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.4077226718266805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,power_law_1.2,1.3657407760620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,balanced,0.030495998760064442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,balanced,0.037104000647862755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,balanced,0.0391893337170283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,balanced,0.0383840004603068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,balanced,0.053216000398000084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,balanced,0.06863999863465627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,balanced,0.0779306689898173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,balanced,0.09923733274141948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,balanced,0.15076800187428793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,balanced,0.18617600202560425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,balanced,0.25277866919835407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,balanced,0.3144800066947937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,power_law_1.01,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,power_law_1.01,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,power_law_1.01,0.07029759883880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,power_law_1.01,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,power_law_1.01,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,power_law_1.01,0.07630079984664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,power_law_1.01,0.07551360130310059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,power_law_1.01,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,power_law_1.01,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,power_law_1.01,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,power_law_1.01,0.09359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,power_law_1.01,0.0953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,power_law_1.01,0.10009599924087524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,power_law_1.01,0.11182080507278443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,power_law_1.01,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,power_law_1.01,0.13791999816894532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,power_law_1.01,0.1572864055633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,power_law_1.01,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,power_law_1.01,0.2373823881149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,power_law_1.01,0.31836800575256347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,power_law_1.01,0.39559040069580076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,power_law_1.01,0.5302976131439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,power_law_1.01,0.7070015907287598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,power_law_1.01,0.06420480012893677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,power_law_1.01,0.0630079984664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,power_law_1.01,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,power_law_1.01,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,power_law_1.01,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,power_law_1.01,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,power_law_1.01,0.08439040184020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,power_law_1.01,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,power_law_1.01,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,power_law_1.01,0.10196479558944702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.01,0.13023359775543214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.01,0.14922239780426025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.01,0.18950400352478028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.01,0.23300480842590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.01,0.3472383975982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.01,0.45769600868225097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.01,0.6637951850891113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.01,0.8831487655639648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.14171520471572877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.14970879554748534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.149452805519104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.16001280546188354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.16965759992599488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.17690240144729613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.18936320543289184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.1965440034866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.20689918994903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.22225279808044435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.2433919906616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.2583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.3165247917175293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.34087040424346926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.4362624168395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.5512256145477294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,power_law_1.01,0.7250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,power_law_1.01,0.8816127777099609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,power_law_1.01,1.2778112411499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,power_law_1.01,1.5780863761901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,balanced,0.09678933024406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,balanced,0.11638399958610535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,balanced,0.14802666505177817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,balanced,0.21313067277272543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,balanced,0.3393333355585734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,balanced,0.5869653224945068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,balanced,0.845034678777059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,balanced,0.840341329574585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,balanced,0.850762685139974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,balanced,0.8537226517995199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,balanced,0.8552266756693522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,balanced,0.8585440317789713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,balanced,0.8716639677683512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,balanced,0.8811840216318766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,balanced,0.8872640132904053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,balanced,0.9004373550415039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,balanced,0.9342666467030843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,balanced,0.9760106404622396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,balanced,1.0329759915669758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.1336053212483723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.1950506369272869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,balanced,1.3700373967488606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,balanced,1.5474665959676106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,balanced,1.8588426907857258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,balanced,2.398101329803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,balanced,3.0616321563720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,balanced,4.218570709228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,power_law_1.01,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,power_law_1.01,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,power_law_1.01,0.0573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,power_law_1.01,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,power_law_1.01,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,power_law_1.01,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,power_law_1.01,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,power_law_1.01,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,power_law_1.01,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.01,0.103603196144104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.01,0.1239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.01,0.14291199445724487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.01,0.18888959884643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.01,0.22821760177612305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.01,0.3053375959396362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.01,0.42021760940551756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.01,0.6546239852905273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.01,0.8479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,balanced,0.039850667119026184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,balanced,0.04322133461634318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,balanced,0.04756266872088114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,balanced,0.05850133299827576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,balanced,0.06238933404286703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,balanced,0.06669333577156067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,balanced,0.08130133152008057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,balanced,0.09749333063761394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,balanced,0.13170133034388223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,balanced,0.16168533762296042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,balanced,0.21458667516708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,balanced,0.2658986647923787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.01,0.07809919714927674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.01,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.01,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.01,0.07282559871673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.01,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.01,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.01,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.01,0.07938560247421264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.01,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.01,0.08341119885444641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.01,0.08925439715385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.01,0.09016960263252258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.01,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.01,0.11367679834365844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.01,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.01,0.13969919681549073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.01,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.01,0.2183232069015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.01,0.2259903907775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.01,0.2892863988876343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.01,0.3541055917739868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.01,0.4722623825073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.01,0.5328192234039306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,power_law_1.2,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,power_law_1.2,0.146560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,power_law_1.2,0.2266752004623413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,power_law_1.2,0.3179327964782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,power_law_1.2,0.40668158531188964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,power_law_1.2,0.6130112171173095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,power_law_1.2,0.6591360092163085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,power_law_1.2,0.6627647876739502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,power_law_1.2,0.6773759841918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,power_law_1.2,0.7041024208068848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,power_law_1.2,0.7346240043640136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,power_law_1.2,0.7440896034240723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,power_law_1.2,0.7875840187072753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,power_law_1.2,0.8212032318115234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,power_law_1.2,0.8373503684997559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,power_law_1.2,0.8690496444702148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,power_law_1.2,0.9420096397399902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,power_law_1.2,1.0113408088684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,power_law_1.2,1.1777600288391112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,power_law_1.2,1.2241408348083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,power_law_1.2,1.4344703674316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,power_law_1.2,1.6793279647827148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,power_law_1.2,2.1548160552978515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,power_law_1.2,2.566054344177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,power_law_1.2,3.4735553741455076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,power_law_1.2,4.370278549194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,power_law_1.2,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,power_law_1.2,0.051712000370025636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,power_law_1.2,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,power_law_1.2,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,power_law_1.2,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,power_law_1.2,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,power_law_1.2,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,power_law_1.2,0.1256767988204956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.2,0.14476159811019898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.2,0.1708351969718933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.2,0.2066495895385742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.2,0.2776576042175293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.2,0.3623680114746094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.2,0.4992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.2,0.6368127822875976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,balanced,0.05841066439946493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,balanced,0.07459199925263722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,balanced,0.09045333663622539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,balanced,0.0962666670481364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,balanced,0.09732799728711446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,balanced,0.09590933720270793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,balanced,0.09821333487828572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,balanced,0.09785067041714986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,balanced,0.09923733274141948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,balanced,0.10133333007494609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,balanced,0.1027786632378896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,balanced,0.10300266742706299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,balanced,0.10876799623171489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,balanced,0.11145066221555074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,balanced,0.11782933274904887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,balanced,0.13153599699338278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,balanced,0.14333333571751913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,balanced,0.18810667594273886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,balanced,0.19172267119089761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,balanced,0.25645333528518677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,balanced,0.2635466655095418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,balanced,0.352783997853597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,balanced,0.4131840070088704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,balanced,0.5777386824289957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,balanced,0.7037813663482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.2,0.9089407920837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.2,1.1969280242919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.15827840566635132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.1547711968421936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.15175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.15096960067749024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.14651520252227784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.14867199659347535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.14275200366973878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.1490048050880432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.1490623950958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.15492479801177977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.1589184045791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.15790079832077025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.15724799633026124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.1660863995552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.15909759998321532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.17430399656295775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.1778048038482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.20122239589691163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.2248703956604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.2850303888320923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.32233600616455077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.4166975975036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.5262976169586182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,power_law_1.01,0.7199359893798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,power_law_1.01,0.9442239761352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.18325120210647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.3001919984817505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.3963263988494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.01,0.6502463817596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,balanced,0.03800000001986822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.01,0.6908927917480469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.01,0.7088064193725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.01,0.7312064170837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.01,0.7427648067474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,balanced,0.041989331444104515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,balanced,0.04610666632652283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.07830399771531422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.09207466244697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.11793599526087443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.1486186683177948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.19549334049224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.23862934112548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.01,0.7829887866973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.01,0.7815487861633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.01,0.8006784439086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.01,0.8278400421142578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.01,0.8657343864440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.01,0.8873472213745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.01,0.9470784187316894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.0625151634216308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.01,1.0912704467773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.01,1.2887807846069337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.01,1.3123647689819335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.01,1.6056831359863282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.01,1.8868352890014648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.01,2.4305919647216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.01,3.0356351852416994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.01,4.214828872680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.01,5.424224090576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,balanced,0.027658666173617046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,balanced,0.029882666965325672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,balanced,0.0363520011305809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,balanced,0.07648533085982005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,balanced,0.09521599610646565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,balanced,0.12641599774360657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,balanced,0.15262933572133383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,power_law_1.01,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,power_law_1.01,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,power_law_1.01,0.10725120306015015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,power_law_1.01,0.11286400556564331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,power_law_1.01,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,power_law_1.01,0.11777280569076538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,power_law_1.01,0.11986559629440308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,power_law_1.01,0.12546559572219848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,power_law_1.01,0.12970240116119386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,power_law_1.01,0.13376640081405639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,power_law_1.01,0.13210879564285277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,power_law_1.01,0.13960319757461548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,power_law_1.01,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,power_law_1.01,0.17292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.1846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.2266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.24487040042877198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.31454079151153563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.37900800704956056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.5116735935211182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.6360191822052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,power_law_1.01,0.8896575927734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.1644672393798827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,balanced,0.028368001182874043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,balanced,0.029717333614826202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,balanced,0.029653333127498627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,balanced,0.038021333515644073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,balanced,0.038949333131313324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,balanced,0.05738666653633118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,balanced,0.06638933221499126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,balanced,0.07298133273919423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,balanced,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,balanced,0.10181867082913716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,balanced,0.13038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,balanced,0.15409599741299948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,balanced,0.21170133352279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,balanced,0.24894932905832926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.047391998767852786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.10006400346755981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.11618560552597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.13717119693756102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.18129279613494872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.24459519386291503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.30227839946746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,power_law_1.01,0.4098944187164307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,power_law_1.01,0.5554815769195557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.2,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.2,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.2,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.2,0.039212799072265624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.2,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.2,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.2,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.2,0.093094402551651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.2,0.11331839561462402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.2,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.2,0.21310079097747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.2,0.23767681121826173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,power_law_1.01,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,power_law_1.01,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,power_law_1.01,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,power_law_1.01,0.08587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,power_law_1.01,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,power_law_1.01,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,power_law_1.01,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,power_law_1.01,0.11076480150222778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,power_law_1.01,0.11980799436569214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,power_law_1.01,0.1253056049346924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,power_law_1.01,0.12636799812316896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,power_law_1.01,0.14410239458084106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,power_law_1.01,0.16601599454879762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,power_law_1.01,0.2007296085357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,power_law_1.01,0.23887360095977783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,power_law_1.01,0.30478720664978026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,power_law_1.01,0.3844671964645386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,power_law_1.01,0.5193600177764892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,power_law_1.01,0.6577407836914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,power_law_1.01,0.9526080131530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,power_law_1.01,1.2383808135986327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,balanced,0.03190933416287104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,balanced,0.03282133241494497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,balanced,0.0383146678407987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,balanced,0.042591998974482216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,balanced,0.05639466643333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,balanced,0.06668266654014587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,balanced,0.06744533280531566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,balanced,0.08624533812204997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,balanced,0.09648000200589497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,balanced,0.12239467104276021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,balanced,0.1595253348350525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,balanced,0.20664000511169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,balanced,0.2662079930305481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,balanced,0.39935465653737384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,balanced,0.5461653470993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,power_law_1.01,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,power_law_1.01,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,power_law_1.01,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,power_law_1.01,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,power_law_1.01,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,power_law_1.01,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,power_law_1.01,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,power_law_1.01,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,power_law_1.01,0.09599999785423279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,power_law_1.01,0.10615040063858032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,power_law_1.01,0.14223999977111818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,power_law_1.01,0.18090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,power_law_1.01,0.2491136074066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,power_law_1.01,0.3549056053161621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,power_law_1.01,0.39857280254364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.1431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.24479999542236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.24265599250793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.23767681121826173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.2346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.18791040182113647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.2091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.2215359926223755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.23123838901519775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.21870720386505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.2225856065750122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.22329599857330323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.23659520149230956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.245849609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.24607999324798585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.23896958827972412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.2646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.29443199634552003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.30117120742797854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.3327552080154419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.3612159967422485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.43277440071105955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.47727360725402834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.5775487899780274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,power_law_1.01,0.6823808193206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,power_law_1.01,0.8980863571166993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,power_law_1.01,1.144991970062256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.10992640256881714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.1510655999183655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.22514560222625732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.23340160846710206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.23556480407714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.24198400974273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.24466559886932374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.2500607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.2506112098693848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.2549184083938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.2634943962097168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.2626944065093994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.282041597366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.3020224094390869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.3242687940597534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.31296000480651853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.355347204208374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.3976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,power_law_1.2,0.47220478057861326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,power_law_1.2,0.5706175804138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,power_law_1.2,0.7492800235748291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,power_law_1.2,0.8936384201049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,power_law_1.2,1.275001621246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,balanced,0.07119999825954437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,balanced,0.09693333506584167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,balanced,0.1471519966920217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,balanced,0.24878400564193726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,balanced,0.34561065832773846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,balanced,0.3460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,balanced,0.3471466700236003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,balanced,0.3485120137532552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,balanced,0.3496053218841553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,balanced,0.35601600011189777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,balanced,0.359007994333903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,balanced,0.3610239823659261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,balanced,0.37067198753356934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,balanced,0.3755040168762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,balanced,0.38701868057250977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,balanced,0.40621864795684814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,balanced,0.4208960135777791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,balanced,0.4554613431294759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,balanced,0.48741332689921063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,balanced,0.5615679820378622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,balanced,0.6329439878463745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,balanced,0.7795039812723795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,balanced,0.9463840325673422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,balanced,1.24453870455424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,balanced,1.5760854085286458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,power_law_1.2,1.641049575805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,power_law_1.2,2.3659008026123045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,balanced,0.09739733735720317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,balanced,0.11970667044321696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,balanced,0.16590399543444315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,balanced,0.2493226726849874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,balanced,0.4054933389027913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,balanced,0.5662613312403361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,balanced,0.5638773441314697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,balanced,0.565829316775004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,balanced,0.5654773314793905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,balanced,0.5678079922993978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,balanced,0.5750240087509155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,balanced,0.572816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,balanced,0.5816373427708944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,balanced,0.5924053192138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,balanced,0.5990826686223348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,balanced,0.6166773239771525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,balanced,0.6594506502151489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,balanced,0.6924533049265543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,balanced,0.7728586991628011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,power_law_1.2,0.06051200032234192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,balanced,0.8418506781260172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,balanced,0.9923786322275797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,balanced,1.1559627056121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,balanced,1.4860906600952148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,power_law_1.2,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,balanced,1.9703946113586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,power_law_1.2,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,balanced,2.5878079732259116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,power_law_1.2,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,balanced,3.49674129486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,power_law_1.2,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,power_law_1.2,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,power_law_1.2,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,power_law_1.2,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,power_law_1.2,0.09290879964828491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,power_law_1.2,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,power_law_1.2,0.10689280033111573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,power_law_1.2,0.10911999940872193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,power_law_1.2,0.1116096019744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,power_law_1.2,0.12958719730377197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.2,0.14163199663162232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.2,0.1684991955757141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.2,0.2101439952850342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.2,0.28065919876098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.2,0.3622656106948853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.2,0.5094592094421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.2,0.7044032096862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.2,0.9468416213989258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.2,1.296947193145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,power_law_1.2,3.110425567626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.01,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.01,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.01,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.01,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.01,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.01,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.01,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.01,0.0929472029209137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.01,0.11400320529937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.01,0.1481152057647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.01,0.20972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.01,0.22479360103607177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,balanced,0.05450133482615153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,balanced,0.08660800258318584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,balanced,0.08931733171145122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,balanced,0.09047466516494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,balanced,0.09088533123334248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,balanced,0.0936906635761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,balanced,0.09485866626103719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,balanced,0.09834667046864827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,balanced,0.10227200388908386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,balanced,0.10455466310183208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,balanced,0.11337066690127055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,balanced,0.12873066465059915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,balanced,0.13237866759300232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,balanced,0.16706132888793945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,balanced,0.18207999070485434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,balanced,0.2604159911473592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,balanced,0.30477333068847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,balanced,0.43886399269104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,balanced,0.5449866851170858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,balanced,0.7914613087972006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,balanced,1.0209279855092366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.11496959924697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.14557440280914308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.16908799409866332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.19011839628219604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.2226815938949585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.20842878818511962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.22426240444183348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.20276479721069335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.1815616011619568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.19553279876708984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.20361599922180176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.20959360599517823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.22780799865722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.23216640949249268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.23228800296783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.2601344108581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.2830399990081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.32487680912017824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.36948480606079104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.4418367862701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.5219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.6329599857330322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,0.8140159606933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.0309311866760253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,1.4961152076721191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,balanced,0.11973866820335388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,balanced,0.11995200316111247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,balanced,0.12218133608500163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,balanced,0.12484799822171529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,balanced,0.12489066521326701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,balanced,0.1246453324953715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,balanced,0.1283146639664968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,balanced,0.13037332892417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,balanced,0.13179199894269308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,balanced,0.13798933227856955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,balanced,0.13927466670672098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,balanced,0.14432000120480856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,balanced,0.16100266575813293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,balanced,0.1759679913520813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,balanced,0.2081013321876526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,balanced,0.24245333671569824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,balanced,0.35361599922180176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,balanced,0.39565332730611164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,balanced,0.5843253135681152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,balanced,0.7114506562550863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,balanced,1.03603196144104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,balanced,1.359066645304362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,balanced,0.055760001142819725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,balanced,0.06930133203665416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,balanced,0.09294399619102478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,balanced,0.09485866626103719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,balanced,0.09687466422716777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,balanced,0.09796266754468282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,balanced,0.09730666875839233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,balanced,0.09956266482671101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,balanced,0.10121599833170573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,balanced,0.09963200489679973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,balanced,0.10391466816266377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,balanced,0.11065600315729777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,balanced,0.11541866262753804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,balanced,0.12468799948692322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.13529066244761148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.15285332997639975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.17684799432754517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.20696532726287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.23374400536219278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.3259146610895793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.3856106599171956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5535306533177694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.6987360318501791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.12195199728012085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.14807679653167724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.18736640214920045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.23445119857788085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.26121599674224855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.28354558944702146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.3405247926712036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.33619840145111085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.3390144109725952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.3457024097442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.34609279632568357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.3719104051589966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.349727988243103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.37449600696563723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.38974719047546386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.4119103908538818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.4268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.4620992183685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.5187967777252197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.554860782623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,0.7231808185577393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,0.7726272106170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,1.0290687561035157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,1.1896832466125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,1.5854911804199219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,2.1869184494018556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.13749120235443116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.06997119784355163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.08506879806518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.10240639448165893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.12350080013275147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.16350719928741456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.18769919872283936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.24977281093597412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.3166656017303467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.4434048175811768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,power_law_1.01,0.5427584171295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,power_law_1.2,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,power_law_1.2,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,power_law_1.2,0.09187840223312378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,power_law_1.2,0.11005439758300781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,power_law_1.2,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,power_law_1.2,0.1459391951560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,power_law_1.2,0.18010239601135253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,power_law_1.2,0.17703039646148683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,power_law_1.2,0.18994560241699218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,power_law_1.2,0.19415040016174318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,power_law_1.2,0.20848639011383058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,power_law_1.2,0.21223039627075196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,power_law_1.2,0.2207808017730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,power_law_1.2,0.2368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,power_law_1.2,0.24364800453186036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,power_law_1.2,0.26657280921936033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,power_law_1.2,0.316595196723938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,power_law_1.2,0.3290496110916138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,power_law_1.2,0.41037440299987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,power_law_1.2,0.45964798927307127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,power_law_1.2,0.5920576095581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,power_law_1.2,0.6989759922027587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,power_law_1.2,0.9219584465026855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,power_law_1.2,1.202892780303955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,power_law_1.2,1.643391990661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,power_law_1.2,2.180966377258301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,power_law_1.2,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,power_law_1.2,0.04288640022277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,power_law_1.2,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,power_law_1.2,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,power_law_1.2,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,power_law_1.2,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,power_law_1.2,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,power_law_1.2,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,power_law_1.2,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,power_law_1.2,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,power_law_1.2,0.09602559804916382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,power_law_1.2,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,power_law_1.2,0.13511680364608764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,power_law_1.2,0.1866368055343628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,power_law_1.2,0.24750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,power_law_1.2,0.3287935972213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,power_law_1.2,0.4178175926208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.054092800617218016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.07619839906692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.10239360332489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.22838399410247803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,power_law_1.2,0.321235203742981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,power_law_1.2,0.4476607799530029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,power_law_1.01,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,power_law_1.01,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,power_law_1.01,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,power_law_1.01,0.11233279705047608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,power_law_1.01,0.15868799686431884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,power_law_1.01,0.20941441059112548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,power_law_1.01,0.23480958938598634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,power_law_1.01,0.2822144031524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,power_law_1.01,0.2939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,power_law_1.01,0.3039743900299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,power_law_1.01,0.3138943910598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,power_law_1.01,0.33207681179046633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,power_law_1.01,0.3446912050247192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,power_law_1.01,0.3657151937484741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,power_law_1.01,0.3819583892822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,power_law_1.01,0.4035776138305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,power_law_1.01,0.4356224060058594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,power_law_1.01,0.5235072135925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,power_law_1.01,0.5373760223388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,power_law_1.01,0.6455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,power_law_1.01,0.7254271984100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,power_law_1.01,0.9137344360351562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,power_law_1.01,1.110598373413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,power_law_1.01,1.4934144020080566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,power_law_1.01,1.9088447570800782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,power_law_1.01,2.585305595397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,power_law_1.01,3.216281509399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.2,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.2,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.2,0.10208640098571778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.2,0.14165760278701783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.2,0.17146240472793578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,balanced,0.037087999284267426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,balanced,0.040576001008351646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,balanced,0.0432533323764801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,balanced,0.05792533357938131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,balanced,0.07047999898592631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,balanced,0.08897599577903748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,balanced,0.10520000259081523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,balanced,0.1381600002447764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,balanced,0.16472533345222473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,balanced,0.22938666741053262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.2,0.2690239906311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,balanced,0.2907680074373881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,balanced,0.4169386625289917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,balanced,0.5418239831924438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.2,0.3398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,power_law_1.2,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,power_law_1.2,0.06347519755363465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,power_law_1.2,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,power_law_1.2,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,power_law_1.2,0.09292799830436707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,power_law_1.2,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,power_law_1.2,0.1199295997619629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,power_law_1.2,0.13048319816589354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,power_law_1.2,0.1634112000465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,power_law_1.2,0.19505280256271362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,power_law_1.2,0.2607935905456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,power_law_1.2,0.3283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,power_law_1.2,0.4818880081176758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,power_law_1.2,0.6041535854339599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.06115840077400207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.05139840245246887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.07651839852333069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.10355199575424194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.1550271987915039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.2473151922225952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.2998464107513428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.4314688205718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.5502655982971192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.01,0.11262719631195069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.01,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.01,0.18617600202560425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.01,0.21604480743408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.01,0.27100799083709715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.01,0.30618240833282473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.01,0.2933311939239502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.01,0.30222721099853517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.01,0.2943487882614136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.01,0.299782395362854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.01,0.31251840591430663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.01,0.3178623914718628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.01,0.3209856033325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.01,0.34092159271240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.01,0.34790399074554446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.01,0.3629631996154785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.01,0.41185917854309084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.01,0.45221757888793945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.01,0.5593855857849122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.01,0.6200384140014649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.01,0.825062370300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.01,0.8994560241699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.01,1.1866239547729491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.01,1.4615039825439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.01,2.16760311126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.01,2.66876163482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,power_law_1.2,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,power_law_1.2,0.10281599760055542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,power_law_1.2,0.10606720447540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,power_law_1.2,0.1247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,power_law_1.2,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,power_law_1.2,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,power_law_1.2,0.11758079528808593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,power_law_1.2,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,power_law_1.2,0.12709759473800658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,power_law_1.2,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,power_law_1.2,0.13030400276184081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,power_law_1.2,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,power_law_1.2,0.15697920322418213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,power_law_1.2,0.17267199754714965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,power_law_1.2,0.18344320058822633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,power_law_1.2,0.23415679931640626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,power_law_1.2,0.24734079837799072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,power_law_1.2,0.32581119537353515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,power_law_1.2,0.4039616107940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,power_law_1.2,0.5272704124450683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,power_law_1.2,0.6681407928466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,power_law_1.2,0.909324836730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,power_law_1.2,1.227500820159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.12440320253372192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.19379839897155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.2352384090423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,power_law_1.2,0.3314176082611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,power_law_1.2,0.4376192092895508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,power_law_1.2,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,power_law_1.2,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,power_law_1.2,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,power_law_1.2,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,power_law_1.2,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,power_law_1.2,0.08826879858970642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,power_law_1.2,0.09169279932975768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,power_law_1.2,0.09344639778137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,power_law_1.2,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,power_law_1.2,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,power_law_1.2,0.11360640525817871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,power_law_1.2,0.11930240392684936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,power_law_1.2,0.136735999584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,power_law_1.2,0.15234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,power_law_1.2,0.19096319675445556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,power_law_1.2,0.2193471908569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,power_law_1.2,0.28619520664215087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.044863998889923096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.09898239970207215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.12553600072860718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.1701248049736023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.20268800258636474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.2586368083953857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,power_law_1.01,0.3517568111419678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,power_law_1.2,0.35331199169158933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.01,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.01,0.10856319665908813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.01,0.11288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.01,0.11598080396652222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.01,0.12223999500274658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.01,0.12189439535140992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.01,0.12795519828796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.01,0.12799359560012818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,power_law_1.2,0.4868480205535889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.01,0.13032959699630736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.01,0.1363263964653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.01,0.140556800365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.01,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.01,0.1638527989387512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.01,0.18264960050582885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.01,0.21924479007720948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.01,0.2319103956222534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.01,0.2874367952346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.01,0.33822081089019773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.01,0.4442431926727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.01,0.5473728179931641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.01,0.7630784034729003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.01,0.9731904029846191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,power_law_1.2,0.6239232063293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,power_law_1.01,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,power_law_1.01,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,power_law_1.01,0.06113280057907104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,power_law_1.01,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,power_law_1.01,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,power_law_1.01,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,power_law_1.01,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,power_law_1.01,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,power_law_1.01,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,power_law_1.01,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,power_law_1.01,0.07333120107650756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,power_law_1.01,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,power_law_1.01,0.11855360269546508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,power_law_1.01,0.13960959911346435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,power_law_1.01,0.17976319789886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,power_law_1.01,0.2285248041152954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,power_law_1.01,0.3045439958572388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,power_law_1.01,0.3900223970413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,power_law_1.2,0.8869952201843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,power_law_1.01,0.5707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,power_law_1.01,0.7316031932830811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,power_law_1.2,1.1348416328430175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.01,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.01,0.121561598777771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.01,0.1859007954597473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.01,0.24038400650024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.01,0.36065919399261476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.01,0.3852992057800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.01,0.399072003364563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.01,0.40857601165771484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.01,0.4174655914306641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.01,0.43031039237976076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.01,0.44040961265563966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.01,0.45690240859985354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.01,0.46759681701660155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.01,0.48503680229187013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.01,0.49873919486999513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.01,0.5363584041595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.01,0.6195903778076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.01,0.6672512054443359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.01,0.8135807991027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.01,0.9768383979797364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.01,1.2300671577453612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.01,1.4377087593078612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.01,1.9616384506225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.01,2.3803264617919924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.01,3.3401344299316404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.01,4.381081771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,power_law_1.01,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,power_law_1.01,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,power_law_1.01,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,power_law_1.01,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,power_law_1.01,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,power_law_1.01,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,power_law_1.01,0.08461440205574036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,power_law_1.01,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,power_law_1.01,0.08518400192260742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,power_law_1.01,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.01,0.11105279922485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.01,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.01,0.14824960231781006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.01,0.19102079868316652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.01,0.22405118942260743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.01,0.3005759954452515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.01,0.3801408052444458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.01,0.5634496212005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.01,0.7354176044464111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.16714880466461182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.27162880897521974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.35450239181518556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.4697408199310303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.6586239814758301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.6388480186462402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.6625343799591065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.6693376064300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.7035136222839355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.7010752201080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.7272895812988281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.7658559799194335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,power_law_1.2,0.8004799842834472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,power_law_1.2,0.8048640251159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,power_law_1.2,0.8332415580749511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,power_law_1.2,0.9389183998107911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,power_law_1.2,0.9360896110534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.0462528228759767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.0574463844299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,power_law_1.2,1.2408896446228028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,power_law_1.2,1.3662336349487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,power_law_1.2,1.6810495376586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,power_law_1.2,2.0337791442871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,balanced,0.09944533308347066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,balanced,0.11538133025169373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,balanced,0.13986666997273764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,balanced,0.19555733601252237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,balanced,0.28627200921376544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,power_law_1.2,2.5120704650878904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,balanced,0.37903467814127606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,balanced,0.3775893449783325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,balanced,0.3811253309249878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,balanced,0.3827466567357381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,balanced,0.3853600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,balanced,0.39530134201049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,balanced,0.3986186583836873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,balanced,0.40338134765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,power_law_1.2,3.1330240249633787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,balanced,0.41122666994730633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,balanced,0.4171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,balanced,0.4334026575088501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,balanced,0.4800479809443156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.5121440092722574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,balanced,0.5908639828364054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,balanced,0.6616373459498087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,balanced,0.8309120337168375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,balanced,0.9933226903279623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,balanced,1.3370347023010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,balanced,1.816970666249593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,balanced,2.502607981363932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,balanced,3.3384265899658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.01,0.04374400079250336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.01,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.01,0.07983360290527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.01,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.01,0.11745920181274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.01,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.01,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.01,0.12177280187606812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.01,0.1262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.01,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.01,0.13152639865875243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.01,0.1381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.01,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.01,0.1578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.01,0.16236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.01,0.1970304012298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.01,0.20095360279083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.01,0.24261760711669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.01,0.2779839992523193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.01,0.3516031980514526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.01,0.44175357818603517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.01,0.5822847843170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.01,0.7046783924102783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,balanced,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,balanced,0.059802666306495667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,balanced,0.08249600231647491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,balanced,0.1204266647497813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,balanced,0.18911999464035034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,balanced,0.18973867098490396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,balanced,0.19373865922292074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,balanced,0.19350934028625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,balanced,0.1943946679433187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,balanced,0.20206934213638306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,balanced,0.21200533707936606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,balanced,0.20853332678476968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,balanced,0.20928533871968588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,balanced,0.21547200282414755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,balanced,0.21819732586542764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,balanced,0.2311519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,balanced,0.25727999210357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,balanced,0.2690986593564351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,balanced,0.3178773323694865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,balanced,0.35369598865509033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,balanced,0.46304531892140705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,balanced,0.5216159820556641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,balanced,0.7230506738026937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,balanced,0.8779093424479166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,balanced,1.2463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,balanced,1.5983254114786785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,balanced,0.02332799881696701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,balanced,0.02385599911212921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,balanced,0.02475733309984207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,balanced,0.025242666403452556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,balanced,0.025733334322770435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,balanced,0.027402666707833607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,balanced,0.02682666728893916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,balanced,0.027317332724730175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,balanced,0.028202667832374573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,balanced,0.02741866558790207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,balanced,0.028666667640209198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,balanced,0.031199999153614044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,balanced,0.03384533276160558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,balanced,0.03643200049797694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,balanced,0.039274667700131737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,balanced,0.04957333207130432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.01,0.028863999247550964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.01,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.01,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.01,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.01,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.01,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.01,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.01,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.01,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.01,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.01,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.01,0.1735103964805603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.01,0.1868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.06038399934768677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.10655360221862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.1590208053588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.19275519847869874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.26067841053009033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.31123199462890627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,power_law_1.01,0.4636352062225342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,power_law_1.01,0.5397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09741439819335937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.11036800146102906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.13541120290756226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.19482879638671874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.2182784080505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.26268160343170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.26727681159973143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.27908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.2910207986831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.30380799770355227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.3088128089904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.32586240768432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.3364032030105591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.338809609413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.3636863946914673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.41829757690429686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.46189441680908205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,0.5618752002716064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,0.6451136112213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,0.8360832214355469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,1.0068608283996583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,1.3743359565734863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.2,1.7350208282470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,balanced,0.05619200070699056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,balanced,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,balanced,0.10550933082898457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,balanced,0.13662933309872946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,balanced,0.13766933480898538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,balanced,0.13803733388582864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,balanced,0.1387999951839447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.2,2.4595903396606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,balanced,0.13821867108345032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,balanced,0.13889599839846292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,balanced,0.1384106675783793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,balanced,0.14270933469136557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,balanced,0.15036267042160034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,balanced,0.15085867047309875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,balanced,0.1591253379980723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,balanced,0.17354132731755575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.1853760083516439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.21242666244506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.23958933353424072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.2945920030275981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.2,3.182899284362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.3524373372395833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.47151466210683185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.6288853486378988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.867194652557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,balanced,1.142031987508138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,power_law_1.2,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,power_law_1.2,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,power_law_1.2,0.028198400139808656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,power_law_1.2,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,power_law_1.2,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,power_law_1.2,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,power_law_1.2,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,power_law_1.2,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.2,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.2,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.2,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.2,0.17566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.2,0.2316351890563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.2,0.344268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.2,0.44583678245544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.2,0.6894144058227539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,power_law_1.01,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,power_law_1.01,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,power_law_1.01,0.14101760387420653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,power_law_1.01,0.15425280332565308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,power_law_1.01,0.18991999626159667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,power_law_1.01,0.21234560012817383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,power_law_1.01,0.24805760383605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,power_law_1.01,0.2505023956298828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,power_law_1.01,0.2554624080657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,power_law_1.01,0.2536767959594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,power_law_1.01,0.25303680896759034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,power_law_1.01,0.27270400524139404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,power_law_1.01,0.26448640823364256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,power_law_1.01,0.2745856046676636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,power_law_1.01,0.28422400951385496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,power_law_1.01,0.2965248107910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,power_law_1.01,0.3017280101776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,power_law_1.01,0.3423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,power_law_1.01,0.33680000305175783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,power_law_1.01,0.37438719272613524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,power_law_1.01,0.4013951778411865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,power_law_1.01,0.47280001640319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,power_law_1.01,0.5065087795257568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,power_law_1.01,0.6265151977539063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,power_law_1.01,0.7403711795806884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,power_law_1.01,0.9831999778747559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,power_law_1.01,1.1789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.2,0.9244223594665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,power_law_1.01,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,power_law_1.01,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,power_law_1.01,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,power_law_1.01,0.06252800226211548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,power_law_1.01,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,power_law_1.01,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,power_law_1.01,0.06643840074539184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,power_law_1.01,0.06952959895133973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,power_law_1.01,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,power_law_1.01,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,power_law_1.01,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,power_law_1.01,0.0771776020526886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,power_law_1.01,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,power_law_1.01,0.10725760459899902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,power_law_1.01,0.13174400329589844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,power_law_1.01,0.15345920324325563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,power_law_1.01,0.20105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,power_law_1.01,0.2452928066253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,power_law_1.01,0.3370752096176147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,power_law_1.01,0.42193918228149413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,power_law_1.01,0.6150911808013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,power_law_1.01,0.8068415641784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,balanced,0.08192533254623413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,balanced,0.1172320048014323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,balanced,0.185263991355896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,balanced,0.32460800806681317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,balanced,0.5938239892323812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,balanced,0.8527893225351969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,power_law_1.2,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,balanced,0.8569066524505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,balanced,0.8647466500600179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,power_law_1.2,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,balanced,0.8669493198394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,balanced,0.8698346614837646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,balanced,0.8790559768676758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,balanced,0.88372802734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,power_law_1.2,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,balanced,0.8933280309041342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,balanced,0.9048639933268229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,balanced,0.9233866532643636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,balanced,0.9349280198415121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,balanced,0.9591840108235677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,balanced,1.13974396387736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,power_law_1.2,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,balanced,1.072218656539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,balanced,1.3788000742594402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,power_law_1.2,0.06542080044746398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,balanced,1.30240003267924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,power_law_1.2,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,balanced,1.684757391611735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,power_law_1.2,0.06799359917640686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,balanced,1.7877333958943684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,balanced,2.6105173428853354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,balanced,2.889871915181478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,balanced,4.3320267995198565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,power_law_1.2,0.0735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,power_law_1.2,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,balanced,5.211989402770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,power_law_1.2,0.09416959881782531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.2,0.11692800521850585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.2,0.14200960397720336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.2,0.16640000343322753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.2,0.21417601108551027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.2,0.2753151893615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.2,0.410265588760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.2,0.5220799922943116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.2,0.737062406539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.2,1.1063551902770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.09242879748344421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.11375999450683594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.13059200048446656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.12251520156860352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.1636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.16083840131759644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.16069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.16005760431289673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.160863995552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.1624127984046936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.16930559873580933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.17985279560089112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.1814144015312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.195251202583313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.21322879791259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.23129599094390868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.27596158981323243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.31113600730895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.3937599897384644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.4617152214050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.6388095855712891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.7903232097625732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.1319744110107421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.4509311676025392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,power_law_1.2,0.07694079875946044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,power_law_1.2,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,power_law_1.2,0.11728639602661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,power_law_1.2,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,power_law_1.2,0.1668992042541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,power_law_1.2,0.1577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,power_law_1.2,0.16370559930801393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,power_law_1.2,0.17006080150604247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,power_law_1.2,0.173363196849823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,power_law_1.2,0.1800447940826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,power_law_1.2,0.18241280317306519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,power_law_1.2,0.18249599933624266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,power_law_1.2,0.18503040075302124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,power_law_1.2,0.19367040395736695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,power_law_1.2,0.21416959762573243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,power_law_1.2,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,power_law_1.2,0.2491584062576294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,power_law_1.2,0.2724096059799194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,power_law_1.2,0.339628791809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,power_law_1.2,0.3749504089355469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,power_law_1.2,0.5045248031616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,power_law_1.2,0.5537983894348144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,power_law_1.2,0.7572415828704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,power_law_1.2,0.9577664375305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,power_law_1.2,1.2733440399169922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,power_law_1.2,1.600351905822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,power_law_1.01,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,power_law_1.01,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,power_law_1.01,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,power_law_1.01,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,power_law_1.01,0.050355201959609984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,power_law_1.01,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,power_law_1.01,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,power_law_1.01,0.05920640230178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.01,0.06904320120811462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.091212797164917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.01,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.01,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.11950080394744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.01,0.10576640367507935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.14929920434951782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.01,0.12962559461593628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.16922880411148072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.01,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.01,0.2181567907333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.21080319881439208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.01,0.3067647933959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.21739521026611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.01,0.3632256031036377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.22169599533081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.23136000633239745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.24547200202941893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.2534143924713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.26287360191345216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.2688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.2878528118133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.3098047971725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.3564863920211792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.4009664058685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.49008641242980955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.5825215816497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.7434368133544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9382143974304199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.2696063995361329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.490015983581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.12042236328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.755142402648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,power_law_1.01,0.07531520128250122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,power_law_1.01,0.07807360291481018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,power_law_1.01,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,power_law_1.01,0.12700159549713136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,power_law_1.01,0.14536960124969484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,power_law_1.01,0.17776639461517335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,power_law_1.01,0.18229759931564332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,power_law_1.01,0.18380800485610962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,power_law_1.01,0.18917759656906127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,power_law_1.01,0.1947711944580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,power_law_1.01,0.2023871898651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,power_law_1.01,0.2045056104660034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,power_law_1.01,0.21187200546264648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,power_law_1.01,0.2218303918838501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,power_law_1.01,0.22795519828796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,power_law_1.01,0.24702720642089843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,power_law_1.01,0.27866239547729493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.01,0.3141247987747192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.01,0.38280320167541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.01,0.44464640617370604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.01,0.580403184890747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.01,0.7085631847381592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.01,0.9835071563720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.01,1.2505151748657226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.01,1.7808319091796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.01,2.3085760116577148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.08783360123634339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.1313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.19340800046920775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.2230207920074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.3369343996047974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.31074559688568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.3324991941452026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.33031039237976073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.3455615997314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.35587201118469236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.3587392091751099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.3683135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.38318719863891604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.38819839954376223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.3988800048828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.41947522163391116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.4602496147155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.497273588180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.595033597946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.6372416019439697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.7905983924865723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,0.9437503814697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,1.2363391876220704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,1.4859264373779297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,2.065017509460449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.027878400683403016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,2.7096384048461912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.08311039805412293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.10663039684295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.13419519662857055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.1975808024406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.199891197681427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.2805567979812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,power_law_1.2,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,power_law_1.2,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,power_law_1.2,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,power_law_1.2,0.09173120260238647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,power_law_1.2,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,power_law_1.2,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,power_law_1.2,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,power_law_1.2,0.11774719953536987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,power_law_1.2,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,power_law_1.2,0.11983360052108764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,power_law_1.2,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,power_law_1.2,0.1286463975906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,power_law_1.2,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,power_law_1.2,0.14570239782333375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,power_law_1.2,0.15421439409255983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,power_law_1.2,0.1767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,power_law_1.2,0.19159040451049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,power_law_1.2,0.23907198905944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,power_law_1.2,0.2762495994567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,power_law_1.2,0.3429759979248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,power_law_1.2,0.41278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,power_law_1.2,0.5587200164794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,power_law_1.2,0.7077311992645263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,power_law_1.2,1.0161151885986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,power_law_1.2,1.28918399810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.1093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.13526400327682495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.1536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.18312959671020507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.185862398147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.19226239919662474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.19966720342636107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.2040640115737915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.21283841133117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.2175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.22157440185546876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.23707520961761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.24053120613098145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.2514879941940308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.27498879432678225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.30770559310913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.34691200256347654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.3972543954849243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.5132863998413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.5959743976593017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.8241408348083497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.0447296142578124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.506771183013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,1.9161151885986327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,power_law_1.2,0.38803839683532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,balanced,0.031109333038330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,balanced,0.05500799914201101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,balanced,0.056847999493281044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,balanced,0.055162668228149414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,balanced,0.05705599983533224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,balanced,0.06378666559855144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,balanced,0.06649599969387054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,balanced,0.08026666442553203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,balanced,0.10143466790517171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,balanced,0.12145066261291504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,balanced,0.14738667011260986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,balanced,0.18386133511861166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,balanced,0.21941866477330527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.08444160223007202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.08112639784812928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.08202880024909973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.1460543990135193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.1755136013031006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.20650238990783693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.2805248022079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.3431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.47124481201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.6047616004943848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,balanced,0.029872000217437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,balanced,0.031178665657838184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,balanced,0.042250668009122215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,balanced,0.046426668763160706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,balanced,0.04660266637802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,balanced,0.05133333305517832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,balanced,0.0599839985370636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,balanced,0.06057600180308024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,balanced,0.08236266672611237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,balanced,0.1058079997698466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,balanced,0.12168533603350322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,balanced,0.16287466883659363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,balanced,0.1953493356704712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,balanced,0.26501333713531494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,balanced,0.3412533203760783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,balanced,0.4904106855392456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,balanced,0.6404213507970175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,power_law_1.2,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,power_law_1.2,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,power_law_1.2,0.1221119999885559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,power_law_1.2,0.1510208010673523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,power_law_1.2,0.16826879978179932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,power_law_1.2,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,power_law_1.2,0.2221951961517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,power_law_1.2,0.2249216079711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,power_law_1.2,0.23788800239562988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,power_law_1.2,0.2511552095413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,power_law_1.2,0.2605952024459839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,power_law_1.2,0.28192639350891113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,power_law_1.2,0.30159358978271483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,power_law_1.2,0.32140800952911375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,power_law_1.2,0.32083840370178224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,power_law_1.2,0.3773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,power_law_1.2,0.42255358695983886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,power_law_1.2,0.5238719940185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,power_law_1.2,0.6050367832183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,power_law_1.2,0.7862143993377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,power_law_1.2,0.9794495582580567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,power_law_1.2,1.319046401977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,power_law_1.2,1.622809600830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,power_law_1.2,2.3038400650024413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,power_law_1.2,2.9748735427856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,power_law_1.2,0.10198400020599366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,power_law_1.2,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,power_law_1.2,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,power_law_1.2,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,power_law_1.2,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,power_law_1.2,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,power_law_1.2,0.0991487979888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,power_law_1.2,0.10344959497451782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,power_law_1.2,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,power_law_1.2,0.11537280082702636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,power_law_1.2,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,power_law_1.2,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,power_law_1.2,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.2,0.14747519493103028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.2,0.1791103959083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.2,0.20140159130096436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.2,0.2599551916122437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.2,0.31203839778900144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.2,0.4337471961975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.2,0.5737343788146972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.2,0.7952191829681396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.2,1.2018176078796388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.053600001335144046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.07316480278968811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.07863680124282837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.09025920033454896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.09575039744377137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.114956796169281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.13652479648590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.17770240306854249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.21696000099182128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.2941312074661255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.3548991918563843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.1328511953353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.13893120288848876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.1393728017807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.14234880208969117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.14391679763793946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.14562560319900514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.1462272047996521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.14855040311813356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.15208959579467773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.15808639526367188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.16071679592132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.18052480220794678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.18502399921417237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.196070396900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.22740480899810792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.2542912006378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3188352108001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,power_law_1.01,0.49699201583862307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.39876480102539064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.5203904151916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,power_law_1.01,0.6388351917266846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.6160704135894776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.8598719596862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.0736767768859863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.5191935539245605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,1.965657615661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.09249280095100403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.09978880286216736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.11411839723587036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.12574080228805543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.13953919410705568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.1522879958152771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.18884479999542236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.22926719188690187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.28088960647583006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,power_law_1.01,0.3499520063400269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,power_law_1.01,0.42803201675415037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,power_law_1.01,0.6331520080566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,power_law_1.01,0.7783423900604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,power_law_1.01,1.267686367034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,balanced,0.10091200470924377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,balanced,0.15264000495274863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,balanced,0.25891733169555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,balanced,0.3092906673749288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,balanced,0.311845342318217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,balanced,0.31328000624974567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,balanced,0.31521066029866535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,balanced,0.31722132364908856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,balanced,0.31754666566848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,balanced,0.3222293257713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,balanced,0.32313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,balanced,0.326474666595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,balanced,0.334175984064738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,balanced,0.34060267607371014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,balanced,0.34913067022959393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,balanced,0.3686346610387166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,balanced,0.38510934511820477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,balanced,0.45508265495300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,balanced,0.46107733249664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,balanced,0.6251200040181478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,balanced,0.6251573165257772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,balanced,0.9007786909739176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,balanced,0.9966026941935221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,balanced,1.3680480321248372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,balanced,1.8010719617207844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,balanced,0.0636053333679835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,balanced,0.07069333394368489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,balanced,0.10157333811124165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,balanced,0.11333333452542622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,balanced,0.15240533153216043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,balanced,0.17308799425760904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,balanced,0.2537493308385213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,balanced,0.33109867572784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,balanced,0.4990933338801066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,balanced,0.6486773490905762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,power_law_1.2,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,power_law_1.2,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,power_law_1.2,0.08112639784812928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,power_law_1.2,0.09796479940414429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,power_law_1.2,0.10204800367355346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,power_law_1.2,0.1050495982170105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,power_law_1.2,0.10765440464019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,power_law_1.2,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,power_law_1.2,0.12053760290145873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,power_law_1.2,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,power_law_1.2,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,power_law_1.2,0.14512640237808228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,power_law_1.2,0.15453439950942993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,power_law_1.2,0.15460480451583863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,power_law_1.2,0.18322559595108032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,power_law_1.2,0.21075201034545898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,power_law_1.2,0.2666304111480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,power_law_1.2,0.3289407968521118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,power_law_1.2,0.430079984664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,power_law_1.2,0.527891206741333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,power_law_1.2,0.7302144050598145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,power_law_1.2,0.9427007675170899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,power_law_1.2,1.3496319770812988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,power_law_1.2,1.7594879150390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.04262399971485138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.06090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.11567360162734985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.15081599950790406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.16145919561386107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.2006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.2656575918197632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,balanced,0.03605333218971888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,balanced,0.04292800029118856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,balanced,0.0598880002895991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,balanced,0.06629333396752675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,balanced,0.0869760016600291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,balanced,0.10714133580525716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.4249983787536621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.1014847993850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.15280640125274658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.2537856101989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.37467520236968993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.5389503955841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.8346559524536132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.8854720115661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,power_law_1.01,0.9479167938232422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,power_law_1.01,0.9719552040100098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,power_law_1.01,0.9918911933898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.0004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.021561622619629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.0445631980895995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.0782143592834472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.0869824409484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.1287551879882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.164896011352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.2729408264160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.311577606201172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.5505855560302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,power_law_1.01,1.6556608200073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,power_law_1.01,1.9654207229614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,power_law_1.01,2.020800018310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,power_law_1.01,2.492736053466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,power_law_1.01,3.2898624420166014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,power_law_1.01,4.219667053222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,power_law_1.01,5.04461441040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.5826879978179932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.10293760299682617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.11813759803771973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.14858880043029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.1901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.24209918975830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.46078081130981446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.5688640117645264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.05915520191192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.11435519456863404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.13138560056686402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.16725120544433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.21502079963684081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.29845120906829836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.37627520561218264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.5695551872253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.7988160133361817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,power_law_1.01,0.0244159996509552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,power_law_1.01,0.0225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,power_law_1.01,0.025312000513076784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,power_law_1.01,0.025964799523353576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,power_law_1.01,0.025286400318145753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,power_law_1.01,0.02655999958515167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,power_law_1.01,0.026419198513031004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,power_law_1.01,0.028249600529670717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,power_law_1.01,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,power_law_1.01,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,power_law_1.01,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,power_law_1.01,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,power_law_1.01,0.04392319917678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.07242879867553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.12446080446243286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.15516159534454346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.1827455997467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,balanced,0.06819200019041698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,balanced,0.09726933638254802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,balanced,0.15005333224932352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,balanced,0.24450133244196573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,balanced,0.3460853497187297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,balanced,0.34678399562835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,balanced,0.3480000098546346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,balanced,0.3476159969965617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,balanced,0.35231467088063556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,balanced,0.356330672899882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,balanced,0.3577866554260254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,balanced,0.3572053511937459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,balanced,0.3666880130767822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,balanced,0.3699359893798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,balanced,0.3771040042241414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,balanced,0.39130667845408124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,balanced,0.4065920114517212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,balanced,0.4321813186009725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,balanced,0.4591253201166789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,balanced,0.514304002126058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,balanced,0.5712693134943644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,balanced,0.6747626463572184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,balanced,0.825429360071818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,balanced,1.0533493359883626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,balanced,1.3291146755218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,power_law_1.01,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,power_law_1.01,0.05726079940795899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,power_law_1.01,0.05932160019874573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,power_law_1.01,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,power_law_1.01,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,power_law_1.01,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,power_law_1.01,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,power_law_1.01,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,power_law_1.01,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,power_law_1.01,0.08492159843444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,power_law_1.01,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,power_law_1.01,0.10881279706954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,power_law_1.01,0.1393728017807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,power_law_1.01,0.1574079990386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,power_law_1.01,0.217305588722229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,power_law_1.01,0.2551680088043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,power_law_1.01,0.3401792049407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,power_law_1.01,0.4667647838592529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.061740797758102414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,balanced,0.08497066299120586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,balanced,0.09875200192133586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,balanced,0.12287466724713643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,balanced,0.19117865959803262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,balanced,0.2576853235562642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,balanced,0.25700799624125165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,balanced,0.2585653265317281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,balanced,0.26098666588465375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,balanced,0.2748853365580241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,balanced,0.25891733169555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,balanced,0.26389867067337036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,balanced,0.28431467215220135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,balanced,0.2642773389816284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,balanced,0.2586933374404907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,balanced,0.2825653354326884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,balanced,0.268010675907135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,balanced,0.27826132376988727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,balanced,0.29097600777943927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,balanced,0.3017973303794861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,balanced,0.3135253389676412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,balanced,0.3402666648228963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,balanced,0.39508267243703205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,balanced,0.5593973398208618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,balanced,0.6208266814549764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,balanced,0.81277863184611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.11515519618988038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.14617600440979003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.16724480390548707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.07906559705734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.08338559865951538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.08680319786071777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.10967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.14222079515457153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.15575679540634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.19351040124893187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.22731521129608154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.30380799770355227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,power_law_1.01,0.6246399879455566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.38229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.559116792678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.6761407852172852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.24599039554595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.3022527933120728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,power_law_1.01,0.8579008102416992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,power_law_1.2,0.36053760051727296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,power_law_1.2,0.5218751907348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,balanced,0.0690880020459493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,balanced,0.09459732969601949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,balanced,0.1442026694615682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,balanced,0.24228266874949136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,balanced,0.4364000161488851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,balanced,0.8245600064595541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,balanced,1.2140639623006184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,balanced,1.2185440063476562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,balanced,1.2178026835123699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,balanced,1.2186986605326335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,balanced,1.2222932974497478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,balanced,1.2326506773630779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,balanced,1.2335573037465413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,balanced,1.238655964533488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,balanced,1.2486080328623455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,balanced,1.2536640167236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,balanced,1.2675466537475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,balanced,1.2940106391906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,balanced,1.3205280303955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,balanced,1.3691199620564778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,balanced,1.4149866104125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,balanced,1.5039893786112468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,balanced,1.5956427256266277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,balanced,1.7816799481709797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,balanced,2.0219786961873374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,balanced,2.3182719548543296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,balanced,3.0664265950520835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.2,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.2,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.2,0.16654720306396484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.2,0.19170559644699098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.2,0.22305920124053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.2,0.24159998893737794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.2,0.24276480674743653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.2,0.2263551950454712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.2,0.23008639812469484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.2,0.23350400924682618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.2,0.24195199012756347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.2,0.24330239295959472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.2,0.2472383975982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.2,0.25795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.2,0.27313919067382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.2,0.28701438903808596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.2,0.30554239749908446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.2,0.359500789642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.2,0.34744958877563475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.2,0.41731839179992675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.2,0.49500160217285155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.2,0.5917568206787109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.2,0.601036787033081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.2,0.7525184154510498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.2,1.0055359840393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.2,1.326700782775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.2,1.5418560028076171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.2,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.2,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.2,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.2,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.2,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.2,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.2,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.2,0.07627519965171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.2,0.0773248016834259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.2,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.2,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.2,0.1309183955192566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.2,0.1477504014968872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.2,0.19615360498428344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.2,0.21341440677642823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.2,0.2962239980697632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.2,0.3722304105758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.2,0.5132991790771484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.2,0.6081920146942139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,2,balanced,0.16532267133394876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,2,balanced,0.1679840087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,2,balanced,0.2190879980723063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,2,balanced,0.32844799757003784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,2,balanced,0.5619413455327352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,2,balanced,0.9977493286132812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,2,balanced,1.0000053246815999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,2,balanced,1.0036479632059734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,2,balanced,1.0031733512878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,2,balanced,1.0083893140157063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,2,balanced,1.0113920370737712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,2,balanced,1.0163146654764812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,2,balanced,1.0200213591257732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,2,balanced,1.0249866644541423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,2,balanced,1.0299999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,2,balanced,1.036410649617513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,2,balanced,1.0467413266499836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,2,balanced,1.0731253623962402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,2,balanced,1.1048266887664795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,2,balanced,1.1651840209960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,2,balanced,1.2163306872049968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,2,balanced,1.3378453254699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,2,balanced,1.4493066469828289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,2,balanced,1.9840373992919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,2,balanced,2.2255093256632485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,2,balanced,3.1101865768432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,2,balanced,4.095088005065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,balanced,0.07736533383528392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,balanced,0.10355200370152791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,balanced,0.1029866635799408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,balanced,0.10342933734258015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,balanced,0.1053546667098999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,balanced,0.10673600435256958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,balanced,0.10548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,balanced,0.10672533512115479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,balanced,0.10818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,balanced,0.10703466335932414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,balanced,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,balanced,0.1092800001303355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,balanced,0.11531733473141988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,balanced,0.12533332904179892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,balanced,0.13141866525014242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,balanced,0.12819733222325644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,balanced,0.15019200245539346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.16548267006874084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.20304532845815024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.24512000878651938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.32427199681599933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.40882134437561035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.5597386757532755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,balanced,0.7254026730855306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.0486773649851482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,balanced,1.366960048675537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,balanced,0.04128533353408178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,balanced,0.029781334102153778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,balanced,0.03256533294916153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,balanced,0.03301866600910822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,balanced,0.032218667368094124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,balanced,0.07976000010967255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,balanced,0.09925333658854167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,balanced,0.12994666894276938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,balanced,0.1556373337904612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,power_law_1.01,0.06511359810829162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,power_law_1.01,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,power_law_1.2,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,power_law_1.01,0.07740799784660339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,power_law_1.2,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,power_law_1.2,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,power_law_1.2,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,power_law_1.2,0.06833919882774353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,power_law_1.2,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.2,0.11346559524536133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.2,0.13214080333709716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.2,0.17356159687042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,power_law_1.01,0.07886719703674316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.2,0.2521984100341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.2,0.3164799928665161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,power_law_1.01,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,power_law_1.01,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,power_law_1.01,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,power_law_1.01,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,power_law_1.01,0.07500799894332885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,power_law_1.01,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,power_law_1.01,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,power_law_1.01,0.07886719703674316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,power_law_1.01,0.08706560134887695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,power_law_1.01,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,power_law_1.01,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,power_law_1.01,0.09440000057220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,power_law_1.01,0.09832320213317872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.01,0.11048320531845093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.01,0.13035520315170288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.01,0.15249279737472535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,power_law_1.01,0.11277439594268798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.01,0.1887488007545471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.01,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.01,0.2164223909378052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.01,0.2934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.01,0.1468608021736145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.01,0.35043840408325194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.01,0.16380159854888915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.01,0.4806784152984619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.01,0.21740798950195311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.01,0.6882880210876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.01,0.268723201751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.01,0.3586751937866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.01,0.4612287998199463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.01,0.6984320163726807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.01,0.8904191970825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.2,0.5340672016143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.2,0.6599808216094971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.2,1.11845121383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.2,1.2312704086303712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.10652799606323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.1500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.2446336030960083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.36767361164093015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.497760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.6019775867462158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.8082367897033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.8558272361755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.8587455749511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.9103872299194335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,power_law_1.2,0.9491583824157714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,power_law_1.2,0.9655296325683593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.0168704032897948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.0503168106079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.0681280136108398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.1422592163085938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.2004608154296874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.3278271675109863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.506060791015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.5437055587768556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,power_law_1.2,1.8199615478515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,power_law_1.2,1.989900779724121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,power_law_1.2,2.430099105834961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,power_law_1.2,2.8533567428588866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,power_law_1.2,3.611648178100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,power_law_1.2,4.625171279907226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,2,balanced,0.1438826620578766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,2,balanced,0.15545066197713217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,2,balanced,0.2109973430633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,2,balanced,0.32791467507680255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,2,balanced,0.5708959897359213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,2,balanced,0.6982399622599283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,2,balanced,0.7021599610646566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,2,balanced,0.7064212958017985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,2,balanced,0.705018679300944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,2,balanced,0.7081333001454672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,2,balanced,0.7098986307779948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,2,balanced,0.7124693393707275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,2,balanced,0.716053326924642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,2,balanced,0.7230239709218343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,2,balanced,0.726085344950358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,2,balanced,0.7304746309916178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,2,balanced,0.742186705271403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,2,balanced,0.7639573415120443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,2,balanced,0.7863893508911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,2,balanced,0.8425227006276449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,2,balanced,0.898314634958903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,2,balanced,1.1877866586049397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,2,balanced,1.296997308731079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,2,balanced,1.7662560145060222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,2,balanced,2.306293328603109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,2,balanced,3.0020265579223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,2,balanced,4.061936060587565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,balanced,0.05619200070699056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,balanced,0.05648533503214518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,balanced,0.056464001536369324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,balanced,0.059434667229652405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,balanced,0.0634986658891042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,balanced,0.06604266663392384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,balanced,0.07169066866238911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,balanced,0.08495466907819112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,balanced,0.11355732878049214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,balanced,0.13803733388582864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,balanced,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,balanced,0.21810666720072427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,balanced,0.30320533116658527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,balanced,0.3818240165710449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,balanced,0.050240000089009605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,balanced,0.0703413337469101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,balanced,0.07132266461849213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,balanced,0.07302399973074596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,balanced,0.07263466715812683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,balanced,0.07400533556938171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,balanced,0.07281599938869476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,balanced,0.07408000032107036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,balanced,0.08084266881148021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,balanced,0.08097066481908162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,balanced,0.08468799789746602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,balanced,0.1111893355846405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,balanced,0.12587199608484903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,balanced,0.16078933080037436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,balanced,0.18930133183797201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,balanced,0.26072533925374347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,balanced,0.3548213243484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,balanced,0.5055466492970785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,balanced,0.5761013428370158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,power_law_1.01,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,power_law_1.01,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,power_law_1.01,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,power_law_1.01,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,power_law_1.01,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,power_law_1.01,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,power_law_1.01,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,power_law_1.01,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,power_law_1.01,0.0641215980052948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,power_law_1.01,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,power_law_1.01,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,power_law_1.01,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,power_law_1.01,0.1374848008155823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,power_law_1.01,0.16873600482940673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,power_law_1.01,0.22164480686187743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,power_law_1.01,0.27752959728240967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,power_law_1.01,0.37798399925231935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,power_law_1.01,0.48449277877807617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.09883520007133484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.10272639989852905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.10444799661636353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.11735039949417114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.13031680583953859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.13224960565567018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.13891199827194214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.1591423988342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.1779263973236084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.21778559684753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.25192320346832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.31999359130859373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.40514559745788575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.5725247859954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.6853184223175048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.01,0.037401598691940305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.01,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.01,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.01,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.01,0.037676799297332766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.01,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.01,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.01,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.01,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.01,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.01,0.13174400329589844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.01,0.15379199981689454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.01,0.22047359943389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.01,0.27648000717163085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.0762816429138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,1.3135744094848634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,power_law_1.01,0.029625600576400755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,power_law_1.01,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,power_law_1.01,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,power_law_1.01,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,power_law_1.01,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,power_law_1.01,0.05888640284538269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,power_law_1.01,0.08047999739646912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,power_law_1.01,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,power_law_1.01,0.12129919528961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,power_law_1.01,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,power_law_1.01,0.20465919971466065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,power_law_1.01,0.274783992767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,power_law_1.2,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,power_law_1.2,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,power_law_1.2,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,power_law_1.2,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,power_law_1.2,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,power_law_1.2,0.07537919878959656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,power_law_1.2,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,power_law_1.2,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.2,0.10577919483184814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.2,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.2,0.1486080050468445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.2,0.20314240455627441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.2,0.26386559009552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.2,0.3915776014328003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.2,0.5323840141296386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.2,0.6685696125030518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.2,0.954195213317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,power_law_1.2,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,power_law_1.2,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,power_law_1.2,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,power_law_1.2,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,power_law_1.2,0.06700159907341004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,power_law_1.2,0.07282559871673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,power_law_1.2,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,power_law_1.2,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,power_law_1.2,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,power_law_1.2,0.09054719805717468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,power_law_1.2,0.1030784010887146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,power_law_1.2,0.11780480146408082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,power_law_1.2,0.14156800508499146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,power_law_1.2,0.16520320177078246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,power_law_1.2,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,power_law_1.2,0.2735680103302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,power_law_1.2,0.3759040117263794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,power_law_1.2,0.540339183807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,power_law_1.01,0.39359359741210936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,power_law_1.2,0.6638336181640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,power_law_1.01,0.47116799354553224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,power_law_1.2,0.8996992111206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,power_law_1.01,0.02746239900588989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,power_law_1.01,0.025766399502754212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,power_law_1.01,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,power_law_1.01,0.023526400327682495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,power_law_1.01,0.025248000025749208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,power_law_1.01,0.026464000344276428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,power_law_1.01,0.026655998826026917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,power_law_1.01,0.027020800113677978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,power_law_1.01,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,power_law_1.01,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,power_law_1.01,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,power_law_1.01,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,power_law_1.01,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.08008959889411926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.1344704031944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.16443519592285155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.24070401191711427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.27664639949798586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,balanced,0.036464000741640724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,balanced,0.04716266691684723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,balanced,0.046816001335779824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,balanced,0.04855466882387797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,balanced,0.05230399966239929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,balanced,0.05436266462008158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,balanced,0.06905599931875865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,balanced,0.07645333309968312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,balanced,0.09299199779828389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,balanced,0.10589866836865743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,balanced,0.14386133352915445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,balanced,0.17204799254735312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,balanced,0.23060266176859537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,balanced,0.2937120000521342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,power_law_1.01,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,power_law_1.01,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,power_law_1.01,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,power_law_1.2,0.0987712025642395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,power_law_1.2,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,power_law_1.2,0.12276480197906495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,power_law_1.01,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,power_law_1.2,0.13417600393295287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,power_law_1.01,0.07876480221748353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,power_law_1.2,0.15203839540481567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,power_law_1.2,0.15149439573287965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,power_law_1.01,0.07871999740600585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,power_law_1.2,0.156876802444458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,power_law_1.01,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,power_law_1.2,0.1578752040863037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,power_law_1.01,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,2,power_law_1.2,0.16359039545059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,power_law_1.01,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,2,power_law_1.2,0.16710400581359863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,2,power_law_1.2,0.17044479846954347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,power_law_1.01,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,2,power_law_1.2,0.17111040353775026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,power_law_1.01,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,2,power_law_1.2,0.17858560085296632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,power_law_1.01,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,2,power_law_1.2,0.18222719430923462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,2,power_law_1.2,0.1891136050224304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,power_law_1.01,0.11614079475402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.01,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.01,0.14503040313720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.01,0.16076799631118774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,2,power_law_1.2,0.20506880283355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.01,0.2047935962677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.2,0.22328319549560546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.01,0.2464576005935669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.2,0.2690176010131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.01,0.34741759300231934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.2,0.3159231901168823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.01,0.4470655918121338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.2,0.39157121181488036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.01,0.6492991924285889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.2,0.48129920959472655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.01,0.9263615608215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.2,0.6797887802124023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.2,0.8457663536071778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.2,1.22740478515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.2,1.6217920303344726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.04476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.08666880130767822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.11031039953231811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.11105920076370239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.1186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.11862399578094482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.12286720275878907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.12914559841156006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.12962559461593628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.1364351987838745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.139027202129364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.15461119413375854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.16469119787216185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.1835904002189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.21628160476684571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.2563776016235352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.31554560661315917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.3964416027069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.5031231880187989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,power_law_1.01,0.7051839828491211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,power_law_1.01,0.8942463874816895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.196326446533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,power_law_1.01,1.4830143928527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,balanced,0.07035733262697856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,balanced,0.07165866593519847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,balanced,0.07035733262697856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,balanced,0.07072000205516815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,balanced,0.0719413310289383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,balanced,0.07121600210666656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,balanced,0.07235200206438701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,balanced,0.07363733152548473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,balanced,0.07711466650168101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,balanced,0.08255999783674876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,balanced,0.08705600102742513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,balanced,0.09146666526794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,balanced,0.10366400082906087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,balanced,0.1111306647459666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,balanced,0.136053333679835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,balanced,0.15454933047294617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,balanced,0.20454400777816772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,balanced,0.2450773318608602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,balanced,0.33902935187021893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,balanced,0.4232106606165568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,power_law_1.2,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,power_law_1.2,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,power_law_1.2,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,power_law_1.2,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,power_law_1.2,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,power_law_1.2,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,power_law_1.2,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.2,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.2,0.14434560537338256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.2,0.23470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.2,0.332096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.2,0.47557759284973145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.2,0.6298111915588379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,power_law_1.01,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,power_law_1.01,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.01,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.01,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.01,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.01,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.01,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.01,0.1717311978340149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.01,0.2059328079223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.01,0.344812798500061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,balanced,0.0276853342851003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,balanced,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,balanced,0.029264000554879505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,balanced,0.03879466652870178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,balanced,0.051114668448766075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,balanced,0.09066133697827657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,balanced,0.1206773320833842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,balanced,0.16364799936612448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,balanced,0.20256000757217407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.01,0.40342397689819337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,power_law_1.2,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,power_law_1.2,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,power_law_1.2,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,power_law_1.2,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,power_law_1.2,0.07868160009384155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,power_law_1.2,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,power_law_1.2,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,power_law_1.2,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,power_law_1.2,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,power_law_1.2,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,power_law_1.2,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,power_law_1.2,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,power_law_1.2,0.06511359810829162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,power_law_1.2,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,power_law_1.2,0.08186879754066467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,power_law_1.2,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,power_law_1.2,0.08501120209693909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,power_law_1.2,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,power_law_1.2,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,power_law_1.2,0.10536960363388062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,power_law_1.2,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,power_law_1.2,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,power_law_1.2,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.2,0.1260607957839966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.2,0.11617920398712159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.2,0.14650880098342894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.2,0.1334015965461731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.2,0.1733504056930542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.2,0.15711359977722167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.2,0.23155200481414795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.2,0.20346240997314452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,balanced,0.049600000182787575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,balanced,0.027488000690937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.2,0.2840192079544067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.2,0.24350719451904296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,balanced,0.031119999786218006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.2,0.4084735870361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.2,0.3774976015090942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,balanced,0.04033066580692927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,balanced,0.039359999199708305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.2,0.5348288059234619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.2,0.447430419921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,balanced,0.052229334910710655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,balanced,0.07046400010585785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,balanced,0.07853866616884868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.2,0.8194432258605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.2,0.676307201385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.2,1.0070464134216308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.2,0.964134407043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.04625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.05998079776763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.13766399621963502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.22151679992675782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.2663039922714233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.3558079957962036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,power_law_1.01,0.4726272106170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,power_law_1.01,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,power_law_1.01,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,power_law_1.01,0.09969279766082764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,power_law_1.2,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,power_law_1.01,0.10842239856719971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,power_law_1.2,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,power_law_1.01,0.11522560119628907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,power_law_1.01,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,power_law_1.01,0.12639360427856444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,power_law_1.2,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,power_law_1.01,0.13458559513092042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,power_law_1.2,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,power_law_1.01,0.1441856026649475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,power_law_1.01,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,power_law_1.01,0.15236480236053468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,power_law_1.01,0.18190720081329345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,power_law_1.01,0.20618879795074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,power_law_1.01,0.262662410736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,power_law_1.2,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,power_law_1.01,0.323526406288147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,power_law_1.01,0.42508797645568847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,power_law_1.01,0.5221888065338135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,power_law_1.01,0.7317567825317383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,power_law_1.01,0.9411392211914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,power_law_1.01,1.353382396697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,power_law_1.01,1.7645183563232423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,power_law_1.2,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,power_law_1.2,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,power_law_1.2,0.10770560503005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,power_law_1.2,0.12785919904708862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,power_law_1.2,0.16798720359802247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,power_law_1.2,0.20222721099853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,power_law_1.2,0.2807807922363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,power_law_1.2,0.3439807891845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,power_law_1.2,0.49957761764526365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,power_law_1.2,0.6591872215270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,power_law_1.01,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,power_law_1.01,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,power_law_1.01,0.08995839953422546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,power_law_1.01,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,power_law_1.01,0.09248639941215515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,power_law_1.01,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,power_law_1.01,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,power_law_1.01,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,power_law_1.01,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,power_law_1.01,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,power_law_1.01,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,power_law_1.01,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,power_law_1.01,0.13738880157470704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,power_law_1.01,0.16004480123519899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,power_law_1.01,0.17944320440292358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,power_law_1.01,0.2364032030105591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,power_law_1.01,0.31235198974609374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,power_law_1.01,0.41582717895507815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,power_law_1.01,0.49797759056091306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,power_law_1.01,0.7108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,power_law_1.01,0.9119935989379883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.2,0.04506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.2,0.02874239981174469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.2,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.2,0.029452800750732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.2,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.2,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.2,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.2,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.2,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.2,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.2,0.12315520048141479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.2,0.1576640009880066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.2,0.20432639122009277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,power_law_1.01,0.05183359980583191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,power_law_1.01,0.06506239771842956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,power_law_1.01,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,power_law_1.01,0.15469440221786498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,power_law_1.01,0.18183679580688478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,power_law_1.01,0.24126720428466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,power_law_1.01,0.249017596244812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,power_law_1.01,0.2610176086425781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,power_law_1.01,0.2662911891937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,power_law_1.01,0.2788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,power_law_1.01,0.2846656084060669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,power_law_1.01,0.29407999515533445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,power_law_1.01,0.3063488006591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,power_law_1.01,0.313753604888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,power_law_1.01,0.32549118995666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,power_law_1.01,0.3438591957092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,power_law_1.01,0.38470399379730225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,power_law_1.01,0.39860479831695556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,power_law_1.01,0.4777088165283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,power_law_1.01,0.5305344104766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,power_law_1.01,0.6695295810699463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,power_law_1.01,0.7940864086151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,power_law_1.01,1.0664192199707032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,power_law_1.01,1.3038016319274903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,power_law_1.01,1.8238399505615235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,power_law_1.01,2.313478469848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,power_law_1.01,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,power_law_1.01,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,power_law_1.01,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,power_law_1.01,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,power_law_1.01,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,power_law_1.01,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,power_law_1.01,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,power_law_1.01,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.01,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.01,0.1354688048362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.01,0.16819839477539061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.01,0.21076478958129882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.01,0.2914112091064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.01,0.3888063907623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.01,0.5756415843963623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,power_law_1.01,0.08088319897651672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,power_law_1.01,0.10266239643096924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,power_law_1.01,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,power_law_1.01,0.14705920219421387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,power_law_1.01,0.1537984013557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,power_law_1.01,0.15788160562515258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,power_law_1.01,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,power_law_1.01,0.17134720087051392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,power_law_1.01,0.18557440042495726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,power_law_1.01,0.19141119718551636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,power_law_1.01,0.20621440410614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,power_law_1.01,0.22647039890289306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,power_law_1.01,0.23829119205474852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,power_law_1.01,0.23508479595184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,power_law_1.01,0.28441600799560546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,power_law_1.01,0.3231935977935791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,power_law_1.01,0.4127039909362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,power_law_1.01,0.4918208122253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,power_law_1.01,0.6609280109405518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,power_law_1.01,0.8240063667297364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,power_law_1.01,1.1393152236938477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,power_law_1.01,1.4506688117980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,power_law_1.01,2.070700836181641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,power_law_1.01,2.7228736877441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.01,0.8814016342163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.01,1.162003231048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,balanced,0.042394667863845825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,balanced,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,balanced,0.08009600142637889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,balanced,0.12336533268292744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,balanced,0.1239520013332367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,balanced,0.1206666628519694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,balanced,0.12295466661453247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,balanced,0.12120532989501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,balanced,0.12614400188128153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,balanced,0.12775466839472452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,balanced,0.1272213359673818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,balanced,0.1296266714731852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,balanced,0.13725866874059042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,balanced,0.13964266578356424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,balanced,0.1474506656328837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,balanced,0.1607200006643931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,balanced,0.1683839956919352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,balanced,0.19320533672968546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,balanced,0.21260799964269003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,balanced,0.277621328830719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,balanced,0.30852800607681274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,balanced,0.42930134137471515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,balanced,0.5217013359069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,balanced,0.7301279703776041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,balanced,0.9524373213450114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,power_law_1.01,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,power_law_1.01,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,power_law_1.01,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,power_law_1.01,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,power_law_1.01,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,power_law_1.01,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,power_law_1.01,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,power_law_1.01,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,power_law_1.01,0.06128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,power_law_1.01,0.07775359749794006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,power_law_1.01,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,power_law_1.01,0.1121791958808899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,power_law_1.01,0.13689600229263305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,power_law_1.01,0.19226880073547364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,power_law_1.01,0.2388223886489868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,power_law_1.01,0.3291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,power_law_1.01,0.4112832069396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,balanced,0.04744000236193339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,balanced,0.04766400158405304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,balanced,0.048394665122032166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,balanced,0.05027199784914652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,balanced,0.06624533236026764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,balanced,0.08506133159001668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.0955573320388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.11486933628718059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.13016000390052795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.17077332735061646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.21116799116134644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.29205334186553955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.37012799580891925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.5334826707839966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.6877760092417399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.01,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.01,0.11893119812011718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.01,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.01,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.01,0.1394368052482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.01,0.14661120176315307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.01,0.1418944001197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.01,0.14372479915618896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.01,0.1423616051673889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.01,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.01,0.1508095979690552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.01,0.14895360469818114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.01,0.15542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.01,0.15849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.01,0.1729920029640198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.01,0.17342079877853395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.01,0.1820672035217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.01,0.20726399421691893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.01,0.23077120780944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.01,0.28425600528717043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.01,0.31256320476531985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.01,0.39329280853271487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.01,0.42671999931335447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.01,0.5791168212890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.01,0.7293312072753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.01,0.9766847610473632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.01,1.2221183776855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,balanced,0.06775466601053874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,balanced,0.0739573339621226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,balanced,0.1129919985930125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,balanced,0.13895466923713684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,balanced,0.20921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,balanced,0.23179733753204346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,balanced,0.33315199613571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,balanced,0.4299413363138835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,balanced,0.6198986768722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,balanced,0.7856960296630859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,power_law_1.01,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,power_law_1.01,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,power_law_1.01,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,power_law_1.01,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,power_law_1.01,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,power_law_1.01,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,power_law_1.01,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,power_law_1.01,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,power_law_1.01,0.11313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,power_law_1.01,0.13284480571746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,power_law_1.01,0.13617279529571533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,power_law_1.01,0.1637312054634094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,power_law_1.01,0.18595839738845826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,power_law_1.01,0.24450559616088868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,power_law_1.01,0.27543680667877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,power_law_1.01,0.38709120750427245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,power_law_1.01,0.482374382019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,power_law_1.01,0.6655488014221191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,power_law_1.01,0.8126144409179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.10634880065917969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.11386239528656006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.13267840147018434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.12861440181732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.143449604511261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.14479999542236327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.15236480236053468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.16059520244598388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.1664512038230896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.1880128026008606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.19265279769897461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.23431038856506348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.26629760265350344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.3383039951324463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.42841601371765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.5559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,power_law_1.01,0.6666944026947021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,power_law_1.01,0.932192039489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,power_law_1.01,1.2549568176269532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,power_law_1.01,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,power_law_1.01,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,power_law_1.01,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,power_law_1.01,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,power_law_1.01,0.07525759935379028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,power_law_1.01,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,power_law_1.01,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,power_law_1.01,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,power_law_1.01,0.09660159945487976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,power_law_1.01,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,power_law_1.01,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,power_law_1.01,0.10955519676208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,power_law_1.01,0.11968640089035035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,power_law_1.01,0.1360576033592224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,power_law_1.01,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,power_law_1.01,0.20401279926300048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,power_law_1.01,0.25172479152679444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,power_law_1.01,0.30588159561157224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,power_law_1.01,0.4080383777618408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,power_law_1.01,0.548095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,power_law_1.01,0.7632512092590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,power_law_1.01,0.9462143898010253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,balanced,0.04773866633574168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,balanced,0.05077333251635233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,balanced,0.07634133100509644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,balanced,0.07850133379300435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,balanced,0.07973333199818929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,balanced,0.08703466256459554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,balanced,0.0881119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,balanced,0.08867733677228291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,balanced,0.08695466319719951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,balanced,0.09945600231488545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,balanced,0.10108799735705058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,balanced,0.10795733332633972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,balanced,0.11341333389282227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,balanced,0.12102400263150533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,balanced,0.13832533359527588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,balanced,0.15219199657440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,balanced,0.1914506753285726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,balanced,0.21821866432825723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,balanced,0.273306667804718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,balanced,0.3367679913838704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,balanced,0.45114131768544513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,balanced,0.5905173222223917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,power_law_1.01,0.04625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,power_law_1.01,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,power_law_1.01,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,power_law_1.01,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,power_law_1.01,0.05968639850616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,power_law_1.01,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,power_law_1.01,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,power_law_1.01,0.12055679559707641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,power_law_1.01,0.13915519714355468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,power_law_1.01,0.18035199642181396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,power_law_1.01,0.22822399139404298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,power_law_1.01,0.3158720016479492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,power_law_1.01,0.41614718437194825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,balanced,0.05457599957784017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,balanced,0.07630933324495952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,balanced,0.1133013367652893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,balanced,0.17622933785120645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,balanced,0.1765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,balanced,0.18074132998784384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,balanced,0.18080532550811768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,balanced,0.17987199624379477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,balanced,0.18364266554514566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,balanced,0.1871253252029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,balanced,0.18685867389043173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,balanced,0.18937599658966064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,balanced,0.1963626742362976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,balanced,0.19670933485031128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,balanced,0.2035306692123413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,balanced,0.21547732750574747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,balanced,0.22682666778564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,balanced,0.25885866085688275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,balanced,0.27562665939331055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,balanced,0.34521599610646564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,balanced,0.3874346812566121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,balanced,0.5059413512547811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,balanced,0.5990666548411051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,balanced,0.8381760120391846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,balanced,1.045413335164388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.09928320050239563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.12725759744644166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.18317439556121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.25615999698638914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.34466559886932374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.546451187133789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5795455932617187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6236544132232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6593984127044678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.6894400119781494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.6856448173522949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7016640186309815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7185728073120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7412223815917969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.7445568084716797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,power_law_1.2,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.7732736110687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.8642047882080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.9363200187683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.0935680389404296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,power_law_1.2,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.2018752098083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.4814463615417481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,power_law_1.2,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,power_law_1.2,0.04779520034790039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.6867456436157227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,2.1718208312988283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,power_law_1.2,0.05932160019874573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,power_law_1.2,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,power_law_1.2,0.07682560086250305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,power_law_1.2,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,power_law_1.2,0.10136959552764893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,2.8159040451049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,power_law_1.2,0.1212928056716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,power_law_1.2,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,power_law_1.2,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,power_law_1.2,0.25753600597381593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,power_law_1.2,0.30821120738983154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,power_law_1.2,0.4376512050628662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,power_law_1.2,0.5796351909637452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,3.7800193786621095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,4.740966415405273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,balanced,0.09411199887593587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,balanced,0.14863999684651694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,balanced,0.2466933329900106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,balanced,0.4410666624704997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,balanced,0.4431360165278117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,balanced,0.4432906707127889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,balanced,0.444757342338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,balanced,0.446453332901001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,balanced,0.45059200127919513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,balanced,0.45341865221659344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,balanced,0.4583573341369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,balanced,0.4617013136545817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,balanced,0.4791040023167928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,balanced,0.4828960100809733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,balanced,0.494917352994283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,balanced,0.5155200163523356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,balanced,0.5280160109202067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,balanced,0.581168015797933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,balanced,0.6171146631240845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,balanced,0.6889973481496176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,balanced,0.7734666665395101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,balanced,0.9728480180104574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,balanced,1.1147039731343586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,balanced,1.567519982655843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,balanced,1.9221866925557454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.2,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.2,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.2,0.10561920404434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.2,0.12067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.2,0.15151360034942626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.2,0.14907519817352294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.2,0.156550395488739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.2,0.15205119848251342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.2,0.15192960500717162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.2,0.15633920431137086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.2,0.16311039924621581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.2,0.16733440160751342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.2,0.17059839963912965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.2,0.18259199857711791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.2,0.1837183952331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.2,0.19219199419021607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.2,0.23128321170806884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.2,0.2494976043701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.2,0.30700159072875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.2,0.3738816022872925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.2,0.4762432098388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.2,0.5401535987854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.2,0.7739264011383057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.2,0.9019200325012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.2,1.2530367851257325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.2,1.7166912078857421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,balanced,0.0444106658299764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,balanced,0.06676800052324931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,balanced,0.09129066268603007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,balanced,0.09483200311660767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,balanced,0.0941439966360728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,balanced,0.09435733159383138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,balanced,0.09601599971453349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,balanced,0.09531199932098389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,balanced,0.09777599573135376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,balanced,0.09913600484530131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,balanced,0.10478933652242024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,balanced,0.10419733325640361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,balanced,0.10624000430107117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,balanced,0.11059733231862386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,balanced,0.11433600385983785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,balanced,0.12618133425712585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,balanced,0.13967999815940857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,balanced,0.17345066865285239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,balanced,0.19373865922292074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,balanced,0.26926400264104206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,balanced,0.342303991317749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,balanced,0.49611198902130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,balanced,0.6531946659088135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,balanced,0.029663999875386555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,balanced,0.03640000025431315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,balanced,0.052373334765434265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,balanced,0.06890666484832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,balanced,0.10057066877683003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,balanced,0.1178559958934784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,balanced,0.15942399700482687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,balanced,0.18518932660420737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,balanced,0.2641013264656067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,balanced,0.33979201316833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,balanced,0.48998932043711346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,balanced,0.6376213232676188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,balanced,0.039120001097520195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,balanced,0.06446399788061778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,balanced,0.06479466458161671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,balanced,0.06286933521429698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,balanced,0.06482133269309998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,balanced,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,balanced,0.06473066906134288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,balanced,0.06554666658242543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,balanced,0.06763733426729839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,balanced,0.06798399984836578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,balanced,0.07371733089288075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,balanced,0.0772266685962677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,balanced,0.08430932958920796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,balanced,0.09988266229629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,balanced,0.10798399647076924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,balanced,0.13412800431251526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,balanced,0.15662399927775064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,balanced,0.2089759906133016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,balanced,0.24471465746561685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,balanced,0.33155733346939087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,balanced,0.4249653418858846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.2,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.2,0.09500799775123596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.2,0.11870080232620239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.2,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.2,0.12655359506607056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.2,0.12157440185546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.2,0.12744319438934326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.2,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.2,0.1275264024734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.2,0.12948479652404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.2,0.13744640350341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.2,0.14158719778060913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.2,0.1461824059486389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.2,0.15518720149993898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.2,0.16186239719390869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.2,0.17066240310668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.2,0.19641599655151368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.2,0.2232127904891968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.2,0.27852160930633546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.2,0.3343168020248413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.2,0.44104962348937987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.2,0.47537918090820314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.2,0.6732351779937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.2,0.8947263717651367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.2,1.1937984466552733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.2,1.5895551681518554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.06407679915428162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.06398079991340637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.09533439874649048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.10434559583663941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.10562560558319092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.11107840538024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.11792639493942261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.12707200050354003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.12981760501861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.14942079782485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.17017600536346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.19747840166091918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.22317440509796144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.287827205657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.35190401077270506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.49139838218688964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.6673471927642822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.8672767639160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.1868672370910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,balanced,0.044138665000597634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,balanced,0.09032533566157024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,balanced,0.08941333492596944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,balanced,0.08903466661771138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,balanced,0.09076799949010213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,balanced,0.0942186713218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,balanced,0.09327466289202373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,balanced,0.09546666344006856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,balanced,0.09844266374905904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,balanced,0.09948800007502238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,balanced,0.10517866412798564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,balanced,0.10683733224868774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,balanced,0.11489066481590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,balanced,0.12761599818865457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.14076266686121622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.17205333709716797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.19138665994008383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.27349332968393963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.32499732573827106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.4664693276087443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.5932053327560425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.866757313410441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,balanced,1.1294986406962078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.2,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.2,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.2,0.09010559916496277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.2,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.2,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.2,0.09969919919967651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.2,0.10652799606323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.2,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.2,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.2,0.1206015944480896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.2,0.12491519451141357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.2,0.1322111964225769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.2,0.15730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.2,0.19990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.2,0.2353343963623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.2,0.2874624013900757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.2,0.34773759841918944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.2,0.4572159767150879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.2,0.5432896137237548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.2,0.8311871528625489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.2,0.9959232330322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,balanced,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,balanced,0.08525866270065308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,balanced,0.08454933762550354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,balanced,0.0872373382250468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,balanced,0.08898133039474487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,balanced,0.11235733826955159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,balanced,0.109525332848231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,balanced,0.11457600196202596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,balanced,0.11926933129628499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,balanced,0.12486933668454488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,balanced,0.12148267030715942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,balanced,0.1246560017267863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,balanced,0.13214932878812155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,balanced,0.13015466928482056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,balanced,0.13121066490809122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,balanced,0.13596799969673157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,balanced,0.14115200440088907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,balanced,0.14793599645296732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,balanced,0.1567146678765615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,balanced,0.16261333227157593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,balanced,0.18330667416254678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,balanced,0.20614933967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,balanced,0.2558133403460185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,balanced,0.33389333883921307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,balanced,0.42559464772542316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,balanced,0.5271306832631429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,power_law_1.2,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,power_law_1.2,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,power_law_1.2,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,power_law_1.2,0.13105920553207398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,power_law_1.2,0.15124479532241822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,power_law_1.2,0.15792640447616577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,power_law_1.2,0.15431679487228395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,power_law_1.2,0.15293439626693725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,power_law_1.2,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,power_law_1.2,0.16689280271530152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,power_law_1.2,0.16589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,power_law_1.2,0.17411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,power_law_1.2,0.18796160221099853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,power_law_1.2,0.19464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,power_law_1.2,0.19855999946594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,power_law_1.2,0.21560959815979003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,power_law_1.2,0.23256959915161132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,power_law_1.2,0.2769984006881714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,power_law_1.2,0.31076478958129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,power_law_1.2,0.37221760749816896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,power_law_1.2,0.42936320304870607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,power_law_1.2,0.548364782333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,power_law_1.2,0.6606336116790772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,power_law_1.2,0.9102911949157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,power_law_1.2,1.2118464469909669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,power_law_1.2,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,power_law_1.2,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,power_law_1.2,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,power_law_1.2,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,power_law_1.2,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,power_law_1.2,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,power_law_1.2,0.06218879818916321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,power_law_1.2,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,power_law_1.2,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,power_law_1.2,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.2,0.10891519784927368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.2,0.1341375946998596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.2,0.15946240425109864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.2,0.2058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.2,0.28266239166259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.2,0.39726719856262205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.2,0.5399231910705566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.2,0.8835776329040528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.2,1.2483776092529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,power_law_1.01,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,power_law_1.01,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,power_law_1.01,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,power_law_1.01,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,power_law_1.01,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,power_law_1.01,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,power_law_1.01,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,power_law_1.01,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,power_law_1.01,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,power_law_1.01,0.10009599924087524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,power_law_1.01,0.11820160150527954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,power_law_1.01,0.1409216046333313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,power_law_1.01,0.16784640550613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,power_law_1.01,0.2398848056793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,power_law_1.01,0.3110912084579468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,power_law_1.01,0.4576064109802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,power_law_1.01,0.6115583896636962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,balanced,0.13361600041389465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,balanced,0.21686400969823202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,balanced,0.40647466977437335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,balanced,0.4040213425954183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,balanced,0.3919573227564494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,balanced,0.3954720099767049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,balanced,0.3930399815241496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,balanced,0.3954240083694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,balanced,0.398037314414978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,balanced,0.39711999893188477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,balanced,0.408463994661967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,balanced,0.4190239906311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,balanced,0.4216586748758952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,balanced,0.4243733485539754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,balanced,0.4728693167368571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,balanced,0.4774506489435832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,balanced,0.4834666649500529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,balanced,0.5537546475728353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,balanced,0.5658986568450928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,balanced,0.6988373597462972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,balanced,0.7698613007863363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,balanced,1.0237226486206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,balanced,1.2782080173492432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,balanced,1.8193492889404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,balanced,2.3997012774149575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,balanced,3.5514399210611978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,balanced,4.7251787185668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.06829439997673034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.0680512011051178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.07377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.09552639722824097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.10207359790802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.13356800079345704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.212556791305542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.27768959999084475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.3961600065231323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,power_law_1.2,0.4904191970825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,power_law_1.2,0.716096019744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,power_law_1.01,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,power_law_1.01,0.04275839924812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,power_law_1.01,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,power_law_1.01,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,power_law_1.01,0.10053119659423829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,power_law_1.01,0.1668544054031372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,power_law_1.01,0.23243520259857178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,power_law_1.01,0.2987071990966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,power_law_1.01,0.45045762062072753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,power_law_1.01,0.5646848201751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.09627519845962525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.1297152042388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.15575040578842164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.19967999458312988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.2498944044113159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.27356159687042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.33523199558258054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.35770881175994873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.3478271961212158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.34961280822753904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.3678783893585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.38530559539794923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.39041919708251954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.4075967788696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.4219967842102051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.43507838249206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.4925119876861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.5352831840515136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.6023295879364013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.6528319835662841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,0.7884543895721435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,0.9424192428588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.1964735984802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.4272000312805175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,1.976051139831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,2.5436479568481447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,power_law_1.2,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,power_law_1.2,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,power_law_1.2,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,power_law_1.2,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.2,0.09811199903488159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.2,0.13581440448760987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.2,0.1591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.2,0.2289344072341919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,balanced,0.05129066606362661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,balanced,0.09324799974759419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,balanced,0.1473973294099172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,balanced,0.24447999397913614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,balanced,0.33887465794881183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,balanced,0.3444746732711792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,balanced,0.34661865234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,balanced,0.3464213212331136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,balanced,0.3482133150100708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,balanced,0.3497120141983032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,balanced,0.3514453172683716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,balanced,0.3482666810353597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,balanced,0.35736533006032306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,balanced,0.3580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,balanced,0.36271464824676514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,balanced,0.37625598907470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,balanced,0.3856053352355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,balanced,0.40777067343393963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,balanced,0.42980265617370605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,balanced,0.46834667523701984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,balanced,0.5079306761423746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,balanced,0.5769333442052206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,balanced,0.6843893527984619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,balanced,0.8213439782460531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,balanced,1.0479893684387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.2,0.29146881103515626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.2,0.39745919704437255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.2,0.5494847774505616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.2,0.875276756286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.2,1.1980031967163085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,balanced,0.046394666035970054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,balanced,0.0768693337837855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,balanced,0.08140266438325246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,balanced,0.08189333478609721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,balanced,0.08313600222269694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,balanced,0.08112533390522003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,balanced,0.08331199983755748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,balanced,0.08558932940165202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,balanced,0.09252267082532246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,balanced,0.09532266855239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,balanced,0.09991466999053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,balanced,0.11595732967058818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,balanced,0.11616533001263936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,balanced,0.14858667055765787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,balanced,0.15930666526158652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,balanced,0.21679999430974325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,balanced,0.2092533310254415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,balanced,0.28275199731191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,balanced,0.32500267028808594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,balanced,0.451909343401591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,balanced,0.5361813306808472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,power_law_1.01,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,power_law_1.01,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,power_law_1.01,0.04476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,power_law_1.01,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,power_law_1.01,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,power_law_1.01,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,power_law_1.01,0.1293503999710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.01,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,balanced,0.041093334555625916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,balanced,0.05719999969005585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,balanced,0.05557866891225179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,balanced,0.06070399781068166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,balanced,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,balanced,0.06411733229955037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,balanced,0.07459733386834462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.01,0.19587839841842652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,balanced,0.08767466743787129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,balanced,0.09727999567985535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,balanced,0.12382933497428894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,balanced,0.1404800017674764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,balanced,0.1911946733792623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,balanced,0.2465333342552185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,balanced,0.364303986231486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,balanced,0.4726666609446208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.01,0.24085121154785155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.01,0.35576319694519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.01,0.4564544200897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.01,0.7459328174591064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,balanced,0.04713066418965658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,balanced,0.07295999924341838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,balanced,0.1551093359788259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,balanced,0.20614933967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,balanced,0.20726933081944784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,balanced,0.21042132377624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,balanced,0.21426665782928467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,balanced,0.21082133054733276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,balanced,0.21489065885543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,balanced,0.21801066398620605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,balanced,0.22185067335764566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,balanced,0.22893865903218588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,balanced,0.23224000136057535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,balanced,0.24799466133117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,balanced,0.2643466591835022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,balanced,0.28119466702143353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,balanced,0.31498666604359943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,balanced,0.3436959981918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,balanced,0.4124693473180135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,balanced,0.4959786732991536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,balanced,0.6377493143081665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,balanced,0.8033866882324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,balanced,1.0867573420206706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,balanced,1.3875734011332195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.01,0.9446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.01,1.3865856170654296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.01,1.812019157409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,power_law_1.01,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,power_law_1.01,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,power_law_1.01,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,power_law_1.01,0.0603007972240448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,power_law_1.01,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,power_law_1.01,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,power_law_1.01,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,power_law_1.01,0.08213120102882385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,power_law_1.01,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,power_law_1.01,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,power_law_1.01,0.09702399969100953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,power_law_1.01,0.11207679510116578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,power_law_1.01,0.13184640407562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,power_law_1.01,0.15199359655380248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,power_law_1.01,0.1886016011238098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,power_law_1.01,0.22561919689178467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,power_law_1.01,0.29441280364990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,power_law_1.01,0.3762687921524048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,power_law_1.01,0.5208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,power_law_1.01,0.6751679897308349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,power_law_1.01,0.9777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,power_law_1.01,1.2634431838989257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,balanced,0.04901866614818573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,balanced,0.0682773341735204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,balanced,0.1013706624507904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,balanced,0.13801067074139914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,balanced,0.14138666788736978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,balanced,0.14422399799029031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,balanced,0.14502933621406555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,balanced,0.14592533310254416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,balanced,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,balanced,0.1541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,balanced,0.15555733442306519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,balanced,0.16210666298866272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,balanced,0.16692266861597696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,balanced,0.1751520037651062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,balanced,0.18650132417678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,balanced,0.21405333280563354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,balanced,0.2347093423207601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,balanced,0.2860320011774699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,balanced,0.34832533200581867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,balanced,0.44788801670074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,balanced,0.5412426789601644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,balanced,0.7829226652781168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,balanced,0.9750933647155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,balanced,1.4369920094807942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,balanced,1.8405067125956218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,balanced,0.03306133300065994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,balanced,0.03728000074625015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,balanced,0.03844266633192698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,balanced,0.03908800085385641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,balanced,0.06762133538722992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,balanced,0.07879466811815898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,balanced,0.09074133634567261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,balanced,0.1083626647790273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.06383360028266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.08462079763412475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.0997439980506897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.12101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.1340224027633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.1614464044570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.20887041091918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.2846656084060669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,power_law_1.01,0.09226239919662475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,power_law_1.01,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,power_law_1.01,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,power_law_1.01,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,power_law_1.01,0.09426559805870056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,power_law_1.01,0.0967743992805481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,power_law_1.01,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,power_law_1.01,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,power_law_1.2,0.36560640335083006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,power_law_1.01,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,power_law_1.01,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,power_law_1.01,0.11667200326919555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,power_law_1.01,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,power_law_1.01,0.12585599422454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.01,0.13919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.01,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.01,0.18460160493850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.01,0.24202239513397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.01,0.28705921173095705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.01,0.3793920040130615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.01,0.4946944236755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.01,0.7185215950012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.01,0.9300352096557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,power_law_1.2,0.5441664218902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,power_law_1.2,0.7458111763000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,power_law_1.2,1.1554752349853517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,power_law_1.2,1.4211584091186524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,balanced,0.049322664737701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,balanced,0.05346133311589559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,balanced,0.050901333491007485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,balanced,0.05114666620890299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,balanced,0.06549866497516632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,balanced,0.06532800197601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,balanced,0.06814399858315785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,balanced,0.07094933092594147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,balanced,0.07740266621112823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,power_law_1.01,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,balanced,0.0886293351650238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,balanced,0.09989333152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,balanced,0.12103466192881267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,balanced,0.1381013294061025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,balanced,0.19382933775583902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,balanced,0.24799466133117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,balanced,0.36533331871032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,balanced,0.47444268067677814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,power_law_1.01,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,power_law_1.01,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,power_law_1.01,0.09684479832649232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,power_law_1.01,0.12268160581588745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,power_law_1.01,0.18120959997177125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,power_law_1.01,0.23049600124359132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,balanced,0.09493333101272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,balanced,0.14729600151379904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,balanced,0.243776003519694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,balanced,0.43956267833709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,balanced,0.4413013458251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,balanced,0.43909335136413574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,balanced,0.4420586824417114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,balanced,0.4397173325220744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,balanced,0.44517866770426434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,balanced,0.44863466421763104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,balanced,0.4504693349202474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,balanced,0.45078933238983154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,balanced,0.4601813157399495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,balanced,0.4603840112686157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,balanced,0.4662933349609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,balanced,0.4811040163040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,balanced,0.4984906514485677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,balanced,0.5230826536814371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,balanced,0.5529013474782308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,balanced,0.603930672009786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,balanced,0.6390186548233032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,balanced,0.8274186452229818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,balanced,0.8421119848887125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,balanced,1.2259999910990398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,balanced,1.3420532544453938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,balanced,0.031066666046778362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,balanced,0.03303466737270355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,balanced,0.03324266771475474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,balanced,0.03708266715208689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,balanced,0.04465599854787191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,balanced,0.0558240016301473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,balanced,0.07452266911665599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,balanced,0.08560533324877422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,balanced,0.10934399565060933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,balanced,0.13354133566220602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,balanced,0.17338667313257852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,balanced,0.20414400100708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,balanced,0.28920533259709674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,balanced,0.3646133343378703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.19259519577026368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.23011200428009032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.35768320560455324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.35148160457611083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.36901121139526366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.3816960096359253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.38579840660095216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.3887871980667114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.3977407932281494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.3931967973709106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.4061439990997314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.4133440017700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.4298111915588379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.45491838455200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.4716671943664551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.5201727867126464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,power_law_1.01,0.5688127994537353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,power_law_1.01,0.6295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,power_law_1.01,0.7581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,power_law_1.01,0.8883647918701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,power_law_1.01,1.167347240447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,power_law_1.01,1.467033576965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,power_law_1.01,2.064454460144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,power_law_1.01,2.646771240234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.01,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,power_law_1.01,3.9377281188964846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.01,0.04037759900093078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.01,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.01,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.01,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.01,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.01,0.11708159446716308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.01,0.14069119691848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.01,0.17829760313034057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.01,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.01,0.30139520168304446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,power_law_1.01,5.124595260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,balanced,0.035904000202814736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,balanced,0.03510399907827377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,balanced,0.039317332208156586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,balanced,0.03963200002908707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,balanced,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,balanced,0.061978667974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,balanced,0.0703893353541692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,balanced,0.08302400012811025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,balanced,0.09856533010800679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.09129599928855896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.09599360227584838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.09341440200805665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.0952127993106842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.11436799764633179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.12072960138320923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.1314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.14927999973297118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.1673087954521179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.205184006690979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.23264000415802003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.30487039089202883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.3573440074920654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,power_law_1.01,0.5373504161834717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,power_law_1.01,0.6833663940429687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,balanced,0.045909335215886436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,balanced,0.049269333481788635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,balanced,0.07733866572380066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,balanced,0.08205866813659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,balanced,0.0824533353249232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,balanced,0.08225599924723308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,balanced,0.0842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,balanced,0.08474666873613994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,balanced,0.08609066406885783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,balanced,0.09108266234397888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,balanced,0.08983467022577922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,balanced,0.0913759966691335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,balanced,0.10155733426411946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,balanced,0.09831466277440389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,balanced,0.11698133746782939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,balanced,0.11333866914113362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,balanced,0.14031466841697693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,balanced,0.1460853318373362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,balanced,0.20674665768941244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,balanced,0.2174933354059855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,balanced,0.2799999912579854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,balanced,0.34672534465789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.09367679953575134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.14220160245895386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.1506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.1865023970603943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.2349951982498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.23146240711212157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.30450561046600344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.28564479351043703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.2917504072189331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.3003200054168701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.31329920291900637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.29726080894470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.2866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.32235519886016845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.32398080825805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.33510398864746094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.3582848072052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.3784447908401489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.440550422668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.47516160011291503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.5666880130767822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,power_law_1.01,0.6475840091705323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,power_law_1.01,0.8134336471557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,power_law_1.01,0.9261568069458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,power_law_1.01,1.2158143997192383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,power_law_1.01,1.5321087837219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.2,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.2,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.2,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.2,0.0852288007736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.2,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.2,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.2,0.10765440464019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.2,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.2,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.2,0.1814463973045349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.2,0.23418240547180175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.2,0.27273600101470946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.2,0.3363647937774658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.11093759536743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.2,0.43301119804382326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.15727360248565675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.20867838859558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.2378943920135498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.2911616086959839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.3053951978683472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.3142591953277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.34179201126098635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.34697599411010743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.3596544027328491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.3775487899780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.2,0.666099214553833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.38519039154052737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.4006336212158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.4589375972747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.49954562187194823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.6083712100982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.6802048206329345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.8473471641540528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.0401408195495605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.3708800315856933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.2,0.8075008392333984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.7542848587036133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.3871936798095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,power_law_1.01,2.9490304946899415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.09102720022201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.1419648051261902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.18172160387039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.2916032075881958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.36382079124450684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.5699007987976075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.6931136131286622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,power_law_1.2,0.06660479903221131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,power_law_1.2,0.09219200015068055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,power_law_1.2,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,power_law_1.2,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,power_law_1.2,0.09644160270690919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,power_law_1.2,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,power_law_1.2,0.10933120250701904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,power_law_1.2,0.11496959924697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,power_law_1.2,0.12089600563049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,power_law_1.2,0.12276480197906495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,power_law_1.2,0.13062399625778198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,power_law_1.2,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.2,0.2186176061630249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.2,0.27675518989562986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.2,0.3267712116241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.2,0.5082752227783203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.2,0.5460544109344483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.2,0.7900352001190185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.2,1.0611583709716796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.2,1.63939208984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.2,2.282361602783203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,power_law_1.2,0.07291520237922669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,power_law_1.2,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,power_law_1.2,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,power_law_1.2,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,power_law_1.2,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,power_law_1.2,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,power_law_1.2,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,power_law_1.2,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,power_law_1.2,0.0847104012966156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,power_law_1.2,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,power_law_1.2,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,power_law_1.2,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,power_law_1.2,0.10476800203323364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,power_law_1.2,0.11242879629135132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.2,0.12368639707565307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.2,0.14433280229568482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.2,0.16458239555358886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.2,0.2187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.2,0.2679935932159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.2,0.38289918899536135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.2,0.5164735794067383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.2,0.8176511764526367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.2,0.9728832244873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,balanced,0.07414933542410533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,balanced,0.07537066439787547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,balanced,0.07538133362929027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,balanced,0.07725333174069722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,balanced,0.07629866898059845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,balanced,0.07682666679223378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,balanced,0.07955199976762135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,balanced,0.07964799801508586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,balanced,0.07870933413505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,balanced,0.08499733606974284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,balanced,0.08724266290664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,balanced,0.09819733103116353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,balanced,0.10939199725786845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,balanced,0.12795199950536093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,balanced,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,balanced,0.16781334082285562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,balanced,0.2148639957110087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,balanced,0.2669919927914937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,balanced,0.35574932893117267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,balanced,0.4458400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,power_law_1.01,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,power_law_1.01,0.13662079572677613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,power_law_1.01,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,power_law_1.01,0.24808320999145508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,power_law_1.01,0.29373440742492674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,power_law_1.01,0.4754432201385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,power_law_1.01,0.4828159809112549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,power_law_1.01,0.5077568054199219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,power_law_1.01,0.5094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,power_law_1.01,0.5300288200378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,power_law_1.01,0.5309247970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,power_law_1.01,0.5524096012115478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,power_law_1.01,0.5558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,power_law_1.01,0.5836480140686036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,power_law_1.01,0.5884863853454589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,power_law_1.01,0.6046847820281982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,power_law_1.01,0.6586175918579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,power_law_1.01,0.728172779083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,power_law_1.01,0.7298687934875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,power_law_1.01,0.8250687599182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,power_law_1.01,0.9737343788146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,power_law_1.01,1.1464703559875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,power_law_1.01,1.2263487815856933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,power_law_1.01,1.5164095878601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,power_law_1.01,2.1122623443603517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,power_law_1.01,2.7250816345214846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,power_law_1.01,3.2387649536132814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,balanced,0.0439626673857371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,balanced,0.08018133540948232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,balanced,0.09967999656995137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,balanced,0.09839466214179993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,balanced,0.09713066617647807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,balanced,0.09938666224479675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,balanced,0.09727999567985535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,balanced,0.10045333703358968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,balanced,0.10100266337394714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,balanced,0.10785067081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,balanced,0.11353066563606262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,balanced,0.11741866668065389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,balanced,0.12759466965993246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,balanced,0.14034133156140646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,balanced,0.1534346640110016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,balanced,0.17563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,balanced,0.20127467314402261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,balanced,0.24874132871627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,balanced,0.3107680082321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,balanced,0.4121119976043701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,balanced,0.5233813524246216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,balanced,0.7289013067881266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,balanced,0.937503973642985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,power_law_1.01,0.0298880010843277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,power_law_1.01,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,power_law_1.01,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,power_law_1.01,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,power_law_1.01,0.12270079851150513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,power_law_1.01,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,power_law_1.01,0.218886399269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,power_law_1.01,0.3443968057632446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,power_law_1.01,0.42766718864440917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,power_law_1.2,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,power_law_1.2,0.0522816002368927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,power_law_1.2,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,power_law_1.2,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,power_law_1.2,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,power_law_1.2,0.06800640225410462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,power_law_1.2,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,power_law_1.2,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,power_law_1.2,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,power_law_1.2,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,power_law_1.2,0.10469759702682495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.2,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.2,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.2,0.15735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.2,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.2,0.2251136064529419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.2,0.30368640422821047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.2,0.3854464054107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.2,0.5415359973907471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.2,0.6679488182067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,balanced,0.03994666785001755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,balanced,0.04052799940109253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,balanced,0.040250666439533234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,balanced,0.04455466568470001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,balanced,0.050154666105906166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,balanced,0.05042133231957754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,balanced,0.05787200232346853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,balanced,0.07313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,balanced,0.09097066521644592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,balanced,0.11569066842397054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,balanced,0.13369599978129068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,balanced,0.1793066660563151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,balanced,0.220960001150767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,balanced,0.3058133323987325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,balanced,0.387664000193278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.10203520059585572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.10386559963226319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.12814719676971437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.153766405582428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.14409600496292113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.1718783974647522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.2324671983718872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.26502399444580077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.3727871894836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.47548160552978513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.6336448192596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,0.8929535865783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.2699007987976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,1.5393152236938477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,balanced,0.05827199916044871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,balanced,0.06816000243028005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,balanced,0.12544533610343933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,balanced,0.1800266702969869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,balanced,0.23670933643976846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,balanced,0.2399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,balanced,0.24101867278416952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,balanced,0.24195200204849243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,balanced,0.24818666776021323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,balanced,0.2553173303604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,balanced,0.26019734144210815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,balanced,0.26571200291315716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,balanced,0.27274133761723834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,balanced,0.281333327293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,balanced,0.29846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,balanced,0.3476320107777913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,balanced,0.38636799653371173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,balanced,0.4608853260676066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,balanced,0.5217013359069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,balanced,0.6666560173034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,balanced,0.8650293350219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,balanced,1.1526347001393635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,balanced,1.5820959409077961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,balanced,2.264117399851481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,balanced,2.933487892150879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,2.440902328491211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.11752320528030395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.1677567958831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.15911680459976196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.1677248001098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.172326397895813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.17887359857559204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.18150399923324584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.18447999954223632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.1913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.20005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.20713601112365723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.2202752113342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.24773120880126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.26169600486755373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.31296639442443847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.3661952018737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.47553281784057616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.6075776100158692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,power_law_1.01,0.7652544021606446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,power_law_1.01,0.9762432098388671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,power_law_1.01,1.3125375747680663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,power_law_1.01,1.616819190979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,balanced,0.060234665870666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,balanced,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,balanced,0.17919999361038208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,balanced,0.3036373257637024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,balanced,0.563808004061381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,balanced,0.5619413455327352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,balanced,0.5623946587244669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,balanced,0.5634666681289673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,balanced,0.5673760175704956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,balanced,0.5725280046463013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,balanced,0.5774293343226115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,balanced,0.5820000171661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,balanced,0.5875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,balanced,0.5991199811299642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,balanced,0.60426131884257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,balanced,0.6184159914652506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,balanced,0.6507946650187174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,balanced,0.6770239671071371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,balanced,0.7411093711853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,balanced,0.8153066635131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,balanced,0.9343840281168619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,balanced,1.0194719632466633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,balanced,1.3376372655232747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,balanced,1.5028427441914876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,balanced,2.163013299306234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,balanced,2.6027520497639975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,power_law_1.2,0.037478399276733396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,power_law_1.2,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,power_law_1.2,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,power_law_1.2,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,power_law_1.2,0.04139519929885864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,power_law_1.2,0.049369600415229795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,power_law_1.2,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.2,0.10358400344848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.2,0.12421760559082032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.2,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.2,0.20787839889526366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.2,0.3259903907775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.2,0.4087423801422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.2,0.6352128028869629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,balanced,0.06794133285681407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,balanced,0.07909333209196727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,balanced,0.08276266853014629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,balanced,0.08306666711966197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,balanced,0.08463467160860698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,balanced,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,balanced,0.09246933460235596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,balanced,0.1021066705385844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,balanced,0.10342933734258015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,balanced,0.1050986647605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,balanced,0.1160586675008138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,balanced,0.12428800264994304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,balanced,0.14482667048772177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,balanced,0.15684800346692404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,balanced,0.19524266322453818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,balanced,0.2225546638170878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,balanced,0.28142933050791424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,balanced,0.3487360080083211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,balanced,0.47785067558288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,balanced,0.6327253182729086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.2,0.8792832374572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.12935680150985718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.1682047963142395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.1977023959159851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.26282880306243894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.36823039054870604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,power_law_1.2,0.5551680088043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,power_law_1.2,0.6194687843322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,balanced,0.1142080028851827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,balanced,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,balanced,0.20317333936691284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,balanced,0.3227146665255229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,balanced,0.5657866795857748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,balanced,1.0397813320159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,balanced,1.5267252922058105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,balanced,1.5220692952473958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,balanced,1.5265332857767742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,balanced,1.5313226381937664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,balanced,1.5392959912618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,balanced,1.546101411183675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,balanced,1.5560372670491536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,balanced,1.5709120432535808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,balanced,1.5794666608174641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,balanced,1.5952906608581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,balanced,1.6113066673278809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,balanced,1.6568959554036458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,balanced,1.7025492986043294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,balanced,1.788826624552409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,balanced,1.8669652938842773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,balanced,2.083258628845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,balanced,2.256336053212484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,balanced,2.607205390930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,balanced,3.2145652770996094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,balanced,3.9461228052775064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,balanced,5.3602720896403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,power_law_1.2,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,power_law_1.2,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,power_law_1.2,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,power_law_1.2,0.05112959742546082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,power_law_1.2,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,power_law_1.2,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.2,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.2,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.2,0.12079360485076904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.2,0.16324479579925538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.2,0.21255040168762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.2,0.3215552091598511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.2,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.2,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.2,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.2,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.2,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.2,0.07904000282287597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.2,0.08304640054702758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.2,0.08979840278625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.2,0.09088640213012696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.2,0.10080640316009522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.2,0.41877121925354005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.2,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.2,0.12065279483795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.2,0.152019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.2,0.18018560409545897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.2,0.23509759902954103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.2,0.2410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.2,0.32249600887298585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.2,0.39728000164031985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.2,0.5818304061889649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.2,0.6797887802124023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.2,0.6420671939849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.2,0.8284031867980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,balanced,0.02938133229811986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,balanced,0.029722665747006733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,2,power_law_1.2,0.3841088056564331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,2,power_law_1.2,0.387609601020813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,2,power_law_1.2,0.517523193359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,2,power_law_1.2,0.6544191837310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,2,power_law_1.2,0.8229632377624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,2,power_law_1.2,1.241049575805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,2,power_law_1.2,1.2592576026916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,2,power_law_1.2,1.319929599761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,2,power_law_1.2,1.3137663841247558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,2,power_law_1.2,1.3763903617858886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,2,power_law_1.2,1.4261631965637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,2,power_law_1.2,1.454099178314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,2,power_law_1.2,1.4726207733154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,2,power_law_1.2,1.509875202178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,2,power_law_1.2,1.5484224319458009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,2,power_law_1.2,1.5310463905334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,2,power_law_1.2,1.6158527374267577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,2,power_law_1.2,1.6555904388427733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.2,1.7211584091186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.2,1.9457855224609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.2,2.217708778381348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.2,2.6949184417724608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.2,3.156800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.2,4.0778240203857425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.2,5.075884628295898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.2,6.776268768310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.2,8.762297821044921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,balanced,0.031744000812371574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,balanced,0.03270400067170461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,balanced,0.03387733300526937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,balanced,0.06919466455777486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,balanced,0.04517866671085358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,balanced,0.06490133206049602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,balanced,0.0677706648906072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,balanced,0.0680213322242101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,balanced,0.0699893335501353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,balanced,0.06986133257548015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,balanced,0.07098133365313213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,balanced,0.07833066582679749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,balanced,0.07735466460386912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,balanced,0.08141866823037465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,balanced,0.08746133248011272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,balanced,0.09166933099428813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,balanced,0.10286933183670044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,balanced,0.11187733213106792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,balanced,0.13286399841308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,balanced,0.15565866231918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,balanced,0.19516799847284952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,balanced,0.23997867107391357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,balanced,0.3165866732597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,balanced,0.39792533715566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,1,power_law_1.01,0.15125759840011596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,1,power_law_1.01,0.1748800039291382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,1,power_law_1.01,0.20559999942779542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,1,power_law_1.01,0.2730432033538818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,balanced,0.035936000446478523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,1,power_law_1.01,0.3469311952590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,1,power_law_1.01,0.4174655914306641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,1,power_law_1.01,0.5586175918579102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,1,power_law_1.01,0.5830592155456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,1,power_law_1.01,0.5821951866149903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,1,power_law_1.01,0.6065023899078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,1,power_law_1.01,0.6295743942260742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,1,power_law_1.01,0.6435200214385987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,1,power_law_1.01,0.6454271793365478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,1,power_law_1.01,0.6552512168884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,1,power_law_1.01,0.6683775901794433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,1,power_law_1.01,0.6847680091857911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,1,power_law_1.01,0.7021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,1,power_law_1.01,0.7415040016174317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.01,0.7841216087341308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.01,0.8510720252990722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.01,0.957158374786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.01,1.1025664329528808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.01,1.2847552299499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.01,1.6760000228881835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.01,2.0577407836914063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.01,2.847436714172363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.01,3.624947357177734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,balanced,0.0650079995393753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,balanced,0.09238933523495992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,balanced,0.10959999759991963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,balanced,0.1483786702156067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,balanced,0.18597867091496786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.2,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.2,0.046387198567390445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.2,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.2,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.2,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.2,0.1167296051979065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.2,0.14307839870452882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.2,0.15800319910049437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.2,0.1879807949066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.2,0.22629120349884033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.2,0.29021439552307127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.2,0.34997119903564455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.2,0.5056767940521241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.2,0.6996160030364991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,balanced,0.05563200016816457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,balanced,0.07891733447710673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,balanced,0.0772266685962677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,balanced,0.07772266864776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,balanced,0.07901866734027863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,balanced,0.08141333361466725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,balanced,0.08005866905053456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,balanced,0.08074666559696198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,balanced,0.09263466795285542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,balanced,0.09498133261998494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,balanced,0.09726933638254802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,balanced,0.10977600018183391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,balanced,0.11931733290354411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,balanced,0.14563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,balanced,0.1588159998257955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,balanced,0.19726399580637613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,balanced,0.2338026762008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,balanced,0.31220799684524536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,balanced,0.3797173500061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,balanced,0.5571733315785726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,balanced,0.7064853509267172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,balanced,0.023269332945346832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,balanced,0.02332266668478648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,balanced,0.02317333221435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,balanced,0.024160000185171764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,balanced,0.02588266630967458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,balanced,0.028768000503381092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,balanced,0.027130665878454845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,balanced,0.027637332677841187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,balanced,0.02939733366171519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,balanced,0.03030933439731598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,balanced,0.030943999687830608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,balanced,0.037263999382654824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,balanced,0.03938133269548416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,balanced,0.05880000193913778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,balanced,0.10981866717338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,balanced,0.13023466865221658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.2,0.1268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.2,0.16101759672164917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.2,0.19918080568313598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.2,0.2498431921005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.2,0.2788991928100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.2,0.39203200340270994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.2,0.39419519901275635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.2,0.38474240303039553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.2,0.3969151973724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.2,0.40773758888244627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.2,0.4131135940551758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.2,0.4270912170410156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.2,0.43666558265686034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.2,0.45821437835693357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.2,0.46611838340759276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.2,0.4921152114868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.2,0.517580795288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.2,0.612883186340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.2,0.5995903968811035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.2,0.6956799983978271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.2,0.7598015785217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.2,0.906931209564209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.2,0.9420160293579102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.2,1.1736384391784669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.2,1.4823360443115234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.2,2.1047935485839844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.2,2.328000068664551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,power_law_1.01,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,power_law_1.01,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,power_law_1.01,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,power_law_1.01,0.07157120108604431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,power_law_1.01,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,power_law_1.01,0.07692800164222717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,power_law_1.01,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,power_law_1.01,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,power_law_1.01,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,power_law_1.01,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,power_law_1.01,0.09653760194778442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,power_law_1.01,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.11445759534835816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.13555840253829957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.15325440168380738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.1905727982521057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.23921279907226561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.30573439598083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.3953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,power_law_1.01,0.544652795791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,power_law_1.01,0.669484806060791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,balanced,0.05463466544946035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,balanced,0.05481599768002828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,balanced,0.056794668237368263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,balanced,0.05842666824658712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,balanced,0.05865600208441416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,balanced,0.06378666559855144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,balanced,0.07351999978224437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,balanced,0.07234133283297221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,balanced,0.08934932947158813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,balanced,0.1288800040880839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,balanced,0.13608533143997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,balanced,0.18216532468795776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,balanced,0.21524266401926676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,balanced,0.29843733708063763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,balanced,0.3823733329772949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,balanced,0.5458666483561198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,balanced,0.6848800182342529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,balanced,0.06417599817117055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,balanced,0.06507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,balanced,0.0662720004717509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,balanced,0.06463466584682465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,balanced,0.06679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,balanced,0.06748266518115997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,balanced,0.06905066470305125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,balanced,0.0738506664832433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,balanced,0.073594664533933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,balanced,0.0863146682580312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,balanced,0.09961600104967754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,balanced,0.1088266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,balanced,0.13778133193651834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,balanced,0.1476479967435201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,balanced,0.19820799430211386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,balanced,0.22244799137115479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,balanced,0.31353066364924115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,balanced,0.3738880157470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,power_law_1.01,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,power_law_1.01,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,power_law_1.01,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,power_law_1.01,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,power_law_1.01,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,power_law_1.01,0.12336640357971192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,power_law_1.01,0.14263039827346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,power_law_1.01,0.20468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,power_law_1.01,0.23116800785064698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,power_law_1.01,0.3471296072006226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,power_law_1.01,0.4329535961151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.10276479721069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.13152639865875243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.1758911967277527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.23603200912475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.2747391939163208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.4379136085510254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,balanced,0.14857066671053568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,balanced,0.24114133914311728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,balanced,0.43718401590983075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,balanced,0.8253493309020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,balanced,1.6052746772766113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,balanced,2.3765172958374023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,balanced,2.3752479553222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,balanced,2.3816000620524087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,balanced,2.385669390360514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,balanced,2.380394617716471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,balanced,2.396794637044271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,balanced,2.3942666053771973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,balanced,2.417365392049154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,balanced,2.4307360649108887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,balanced,2.437562624613444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,balanced,2.452693303426107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,balanced,2.476016044616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,balanced,2.5100107192993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,balanced,2.5675199826558432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,balanced,2.6348533630371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,balanced,2.766794522603353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,balanced,2.8641974131266275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,balanced,3.0406932830810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,balanced,3.3372265497843423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,balanced,3.7031679153442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,balanced,4.928826649983724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,balanced,0.04403733213742574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,balanced,0.0705386648575465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,balanced,0.10237333178520203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,balanced,0.15607999761899313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,balanced,0.18412800629933676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,balanced,0.185754656791687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,balanced,0.187717338403066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,balanced,0.18805867433547974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,balanced,0.18862400452295938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,balanced,0.18990933895111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,balanced,0.19301333030064902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,balanced,0.19548799594243368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,balanced,0.19847466548283896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,balanced,0.2079626719156901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,balanced,0.21124267578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,balanced,0.2192479968070984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,balanced,0.244053324063619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,balanced,0.2563626567522685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,balanced,0.30662933985392254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,balanced,0.33563733100891113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,balanced,0.455456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,balanced,0.4954933325449626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,balanced,0.7563786506652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,balanced,0.8171626726786295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,balanced,1.1998079617818196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,balanced,1.4786346753438313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,power_law_1.2,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,power_law_1.2,0.07080320119857789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,power_law_1.2,0.09269760251045227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,power_law_1.2,0.11438720226287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,power_law_1.2,0.11367679834365844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,power_law_1.2,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,power_law_1.2,0.1193343997001648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,power_law_1.2,0.12290560007095337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,power_law_1.2,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,power_law_1.2,0.13411840200424194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,power_law_1.2,0.14320640563964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,power_law_1.2,0.14978560209274291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,power_law_1.2,0.15241600275039674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,power_law_1.2,0.17127039432525634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,power_law_1.2,0.18616960048675538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,power_law_1.2,0.23545598983764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,power_law_1.2,0.24311680793762208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,power_law_1.2,0.33743360042572024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,power_law_1.2,0.3780672073364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,power_law_1.2,0.5199168205261231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,power_law_1.2,0.6522175788879394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,power_law_1.2,0.9450943946838379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,power_law_1.2,1.1353599548339843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,power_law_1.01,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,power_law_1.01,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,power_law_1.01,0.08912000060081482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,power_law_1.01,0.09219840168952942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,power_law_1.01,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,power_law_1.01,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,power_law_1.01,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,power_law_1.01,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,power_law_1.01,0.10616960525512695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,power_law_1.01,0.11889920234680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,power_law_1.01,0.13596800565719605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,power_law_1.01,0.13608319759368898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,power_law_1.01,0.1485759973526001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,power_law_1.01,0.19048320055007933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,power_law_1.01,0.2657984018325806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,power_law_1.01,0.35691521167755125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,power_law_1.01,0.40429439544677737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,power_law_1.01,0.5604991912841797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,power_law_1.01,0.7498496055603028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,power_law_1.2,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,power_law_1.2,0.0754751980304718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,power_law_1.2,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,power_law_1.2,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,power_law_1.2,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,power_law_1.2,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,power_law_1.2,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,power_law_1.2,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,power_law_1.2,0.09608319997787476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,power_law_1.2,0.12057600021362305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,power_law_1.2,0.1386304020881653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,power_law_1.2,0.16921600103378295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,power_law_1.01,1.2083392143249512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,power_law_1.2,0.19005440473556517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,power_law_1.2,0.23571839332580566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,power_law_1.2,0.30421760082244875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,power_law_1.01,1.4853440284729005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,power_law_1.2,0.40964479446411134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,power_law_1.2,0.4688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,power_law_1.01,2.1257535934448244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,power_law_1.2,0.6679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.1146496057510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.11666560173034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.11741440296173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.12423039674758911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1238144040107727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.12670719623565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.12903039455413817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.1333567976951599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.15357439517974852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.1567296028137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.17841919660568237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.1992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.24343678951263428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,power_law_1.01,2.746329689025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.28424320220947263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,power_law_1.2,1.003321647644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.36859519481658937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,power_law_1.01,0.44451198577880857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,power_law_1.01,0.6174975872039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,power_law_1.01,0.7786752223968506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,power_law_1.2,1.2742527961730956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,power_law_1.01,1.1066176414489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,power_law_1.01,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,power_law_1.01,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,power_law_1.01,0.027859199047088622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,power_law_1.01,0.028646400570869444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,power_law_1.01,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,power_law_1.01,0.030656000971794127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,power_law_1.01,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,power_law_1.01,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,power_law_1.01,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,power_law_1.01,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,power_law_1.01,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,power_law_1.01,0.144377601146698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,power_law_1.01,0.20134398937225342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,power_law_1.01,0.2652992010116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,power_law_1.01,1.407027244567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,power_law_1.2,0.04510720074176788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,power_law_1.2,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,power_law_1.2,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,power_law_1.2,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,power_law_1.2,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,power_law_1.2,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,power_law_1.2,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,power_law_1.2,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,power_law_1.2,0.06596480011940002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,power_law_1.2,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,power_law_1.2,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,power_law_1.2,0.13148800134658814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,power_law_1.2,0.15616639852523803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,power_law_1.2,0.22123520374298095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,power_law_1.2,0.2702784061431885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,power_law_1.2,0.39166719913482667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,power_law_1.2,0.5090816020965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,balanced,0.04622933268547058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,balanced,0.07134933272997539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,balanced,0.09963200489679973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,balanced,0.19855999946594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,balanced,0.1988853414853414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,balanced,0.19996267557144165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,balanced,0.199072003364563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,balanced,0.2032960057258606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,balanced,0.204912006855011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,balanced,0.2060640056927999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,balanced,0.20893865823745728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,balanced,0.21504533290863037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,balanced,0.21809067328770956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,balanced,0.2235893408457438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,balanced,0.23518933852513632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,balanced,0.24382400512695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,balanced,0.26950933535893756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,balanced,0.29081066449483234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,balanced,0.3339466651280721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,balanced,0.3821013371149699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,balanced,0.46988264719645184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,balanced,0.5740693410237631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,balanced,0.7398826281229655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,balanced,0.9446240266164144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,balanced,0.040218666195869446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,balanced,0.041989331444104515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,balanced,0.04956266780694326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.058965335289637245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.07632000247637431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.07900799810886383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.09797867139180501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.1418560047944387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.160453329483668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.21964800357818604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.2765386700630188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,balanced,0.06459733347098033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,balanced,0.0865066647529602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,balanced,0.13040000200271606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,balanced,0.19964800278345743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,balanced,0.2587520082791646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,balanced,0.25783999760945636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,balanced,0.25919467210769653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,balanced,0.2608480056126912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,balanced,0.2614399989446004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,balanced,0.2642666697502136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,balanced,0.2664960026741028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,balanced,0.2691359917322795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,balanced,0.272053341070811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,balanced,0.2803200085957845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,balanced,0.2893279989560445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,balanced,0.2938399910926819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,balanced,0.31430933872858685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,balanced,0.33215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,balanced,0.4188266595204671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,balanced,0.4167146682739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,balanced,0.5935680071512858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,balanced,0.5655839840571085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,balanced,0.8189333279927572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,balanced,0.9042932987213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,balanced,1.3336960474650066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,balanced,1.5475200017293294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,power_law_1.2,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,power_law_1.2,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,power_law_1.2,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,power_law_1.2,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,power_law_1.2,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,power_law_1.2,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,power_law_1.2,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,power_law_1.2,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,power_law_1.2,0.11330560445785523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,power_law_1.2,0.1500607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,power_law_1.2,0.20417280197143556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,balanced,0.0902880032857259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,balanced,0.09136000275611877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,balanced,0.09339732925097148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,balanced,0.09315733114878337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,balanced,0.09371733665466309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,balanced,0.09327466289202373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,balanced,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,balanced,0.09682133793830872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,balanced,0.09905067086219788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,balanced,0.10621866583824158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,balanced,0.10750400026639302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,balanced,0.11148800452550252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,balanced,0.12468266487121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,balanced,0.13589866956075033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,balanced,0.15619200468063354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,balanced,0.1739679972330729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,balanced,0.2339359919230143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,balanced,0.26919466257095337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,balanced,0.3848319848378499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,balanced,0.4691093365351359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,balanced,0.6781919797261556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,balanced,0.8930293718973795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,power_law_1.2,0.2643392086029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,power_law_1.2,0.3845247983932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,power_law_1.01,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,power_law_1.01,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,power_law_1.01,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,power_law_1.01,0.09905279874801635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,power_law_1.01,0.10857599973678589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,power_law_1.01,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,power_law_1.01,0.11056640148162841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,power_law_1.01,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,power_law_1.01,0.1067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,power_law_1.01,0.11612160205841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,power_law_1.01,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,power_law_1.01,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,power_law_1.01,0.12684799432754518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,power_law_1.01,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,power_law_1.01,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,power_law_1.01,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.01,0.15292799472808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.01,0.17976959943771362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.01,0.20945920944213867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.01,0.2738048076629639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.01,0.3484863996505737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.01,0.4814335823059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.01,0.61910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.01,0.8843456268310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.01,1.126150417327881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,balanced,0.04450666904449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,balanced,0.04859733581542969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,balanced,0.06917333106199901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,balanced,0.07779733339945476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.10012800494829814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.12407466769218445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.15652267138163248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.1773866613705953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.23658132553100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.2848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.40380267302195233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.5153599977493286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.2,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.2,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.2,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.2,0.056652802228927615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.2,0.060703998804092406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.2,0.06678400039672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.2,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.2,0.12446080446243286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.2,0.1514240026473999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.2,0.19571199417114257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.2,0.20404479503631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.2,0.29755520820617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.2,0.3280128002166748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.2,0.46858878135681153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.2,0.6253568172454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,balanced,0.07032533486684163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,balanced,0.07308800021807353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,balanced,0.0726506660381953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,balanced,0.07644799848397572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,balanced,0.07628799974918365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,balanced,0.0825973351796468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,balanced,0.08621866504351298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,balanced,0.09367466966311137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,balanced,0.09620267152786255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,balanced,0.10914666453997295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,balanced,0.12191999951998393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,balanced,0.14017599821090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,balanced,0.16051200032234192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,balanced,0.20538665850957236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,balanced,0.24385066827138266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,balanced,0.3386346499125163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,balanced,0.418778657913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,balanced,0.05939733485380808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,balanced,0.06250133117039998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,balanced,0.062650665640831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,balanced,0.06253866851329803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,balanced,0.07145066559314728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,balanced,0.06880533198515575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,balanced,0.07681599756081899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,balanced,0.09763200084368388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,balanced,0.11095999677975972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,balanced,0.13037332892417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,balanced,0.15415466825167337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,balanced,0.208624005317688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,balanced,0.25045865774154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,balanced,0.36106133460998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,balanced,0.45790934562683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,balanced,0.6719040075937907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,balanced,0.8764586448669434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,power_law_1.2,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,power_law_1.2,0.1357375979423523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,power_law_1.2,0.13715840578079225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,power_law_1.2,0.14506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,power_law_1.2,0.14587520360946654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,power_law_1.2,0.14672640562057496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,power_law_1.2,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,power_law_1.2,0.15971839427947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,power_law_1.2,0.15989760160446168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,power_law_1.2,0.1675968050956726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,power_law_1.2,0.16919039487838744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,power_law_1.2,0.18216960430145263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,power_law_1.2,0.19514240026474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,power_law_1.2,0.19875199794769288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,power_law_1.2,0.23797121047973632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,power_law_1.2,0.2564608097076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,power_law_1.2,0.32926719188690184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,power_law_1.2,0.3831808090209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,power_law_1.2,0.44237442016601564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,power_law_1.2,0.6587456226348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.09043200016021728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,power_law_1.2,0.8366399765014648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.11673599481582642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.13931519985198976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.17601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.2123647928237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,power_law_1.2,1.052076816558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.29463040828704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,power_law_1.2,1.3860032081604003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.34880640506744387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,power_law_1.2,1.8795007705688476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.48817920684814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,power_law_1.2,2.777779197692871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.7327360153198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,balanced,0.07134933272997539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,balanced,0.10087999701499939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,balanced,0.1567146678765615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,balanced,0.1607200006643931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,balanced,0.16078933080037436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,balanced,0.16242667039235434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,balanced,0.1651946703592936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,balanced,0.16425599654515585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,balanced,0.16586666305859885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,balanced,0.16715733210245767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,balanced,0.1694613297780355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,balanced,0.1721973419189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,balanced,0.17415465911229452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,balanced,0.18242132663726807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,balanced,0.19462400674819946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,balanced,0.19709867238998413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,balanced,0.20691200097401938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,balanced,0.25486934185028076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.274944007396698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.3473333517710368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.4312320152918498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.5855733156204224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.7446773052215576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.0416959921518962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.3617493311564128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,balanced,1.962517261505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,balanced,2.546245257059733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.1223936080932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.4868927955627442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,balanced,0.04004266609748205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,balanced,0.0373279998699824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,balanced,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,balanced,0.04842133323351542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.06450133522351582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.08007466793060303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.09470400214195251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.11840533216794331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.131632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.17840532461802164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.21571199099222818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,power_law_1.01,0.029388800263404846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,power_law_1.01,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,power_law_1.01,0.0304639995098114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,power_law_1.01,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,power_law_1.01,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.01,0.08664960265159607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.01,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.01,0.1375488042831421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.01,0.18742400407791138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.073580801486969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.01,0.21056001186370848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.11027840375900269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.12346240282058715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.01,0.3111743927001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.14051200151443483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.19267840385437013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.01,0.4340991973876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.01,0.6316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.25102720260620115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.01,0.9789055824279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.3128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.2,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.2,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.2,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.2,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.2,0.03203839957714081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.2,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.2,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.2,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.2,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.2,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.2,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.2,0.07576320171356202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.2,0.09768319725990296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.2,0.13715840578079225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,power_law_1.2,0.4073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.2,0.1584768056869507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.2,0.24325120449066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.2,0.2537087917327881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,power_law_1.2,0.5585855960845947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,balanced,0.06088533500830332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,balanced,0.06283733248710632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,balanced,0.06892266869544983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,balanced,0.09075199564297994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,balanced,0.08834133545557658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,balanced,0.08932800094286601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,balanced,0.09490133325258891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,balanced,0.09407466650009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,balanced,0.1011840005715688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,balanced,0.09442133704821269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,balanced,0.0972160001595815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,balanced,0.10405332843462627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,balanced,0.10946666200955708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,balanced,0.1079360047976176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,balanced,0.12010133266448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,balanced,0.12145066261291504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,balanced,0.1386560002962748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,balanced,0.15435733397801718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,balanced,0.18979199727376303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,balanced,0.2241013248761495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,balanced,0.3049386739730835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,balanced,0.3646186590194702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,balanced,0.5143893162409464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,balanced,0.6560159921646118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,power_law_1.01,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,power_law_1.01,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,power_law_1.01,0.0834496021270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,power_law_1.01,0.10674560070037842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,power_law_1.01,0.11400320529937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,power_law_1.01,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,power_law_1.01,0.11461759805679321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,power_law_1.01,0.12023680210113526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,power_law_1.01,0.12016639709472657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,power_law_1.01,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,power_law_1.01,0.13665920495986938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,power_law_1.01,0.14338560104370118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,power_law_1.01,0.15275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,power_law_1.01,0.17374720573425292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,power_law_1.01,0.19028480052948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,power_law_1.01,0.22334721088409423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,power_law_1.01,0.25770881175994875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,power_law_1.01,0.3342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,power_law_1.01,0.4127808094024658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,power_law_1.01,0.5198592185974121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,power_law_1.01,0.6588736057281495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,power_law_1.01,0.9655679702758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,power_law_1.01,1.2636735916137696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.04782719910144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.09336959719657897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.10009599924087524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.1316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.17503999471664428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.25601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.31449599266052247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,balanced,0.029343999922275543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,balanced,0.03277866790692011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,balanced,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,balanced,0.06625600159168243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,balanced,0.09438400467236836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,balanced,0.11940800150235494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,balanced,0.1618826687335968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,balanced,0.19459199905395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,balanced,0.26609599590301514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,balanced,0.3306986689567566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.2,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.2,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.2,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.2,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.2,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.2,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.2,0.06168320178985596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.2,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.2,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.2,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.2,0.09967359900474548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.2,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.2,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.2,0.16729600429534913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.2,0.216812801361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.2,0.2730623960494995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.2,0.35964159965515136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.2,0.477459192276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,power_law_1.01,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,power_law_1.01,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,power_law_1.01,0.054099202156066895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,power_law_1.01,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,power_law_1.01,0.06106879711151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,power_law_1.01,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,power_law_1.01,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,power_law_1.01,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,power_law_1.01,0.08387200236320495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,power_law_1.01,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.01,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.01,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.01,0.1393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.01,0.18192000389099122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.01,0.21298561096191407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.01,0.2729151964187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.01,0.3347264051437378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.01,0.5050623893737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.01,0.6104383945465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,power_law_1.2,0.02805759906768799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,power_law_1.2,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,power_law_1.2,0.02730880081653595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,power_law_1.2,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,power_law_1.2,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,power_law_1.2,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,power_law_1.2,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,power_law_1.2,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,power_law_1.2,0.1461632013320923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,power_law_1.2,0.24121599197387694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,balanced,0.04298666616280874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,balanced,0.0637600024541219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,balanced,0.06531733274459839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,balanced,0.0658079981803894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,balanced,0.06458666423956554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,balanced,0.0665226678053538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,balanced,0.06622399886449178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,balanced,0.0670773337284724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,balanced,0.07009066641330719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,balanced,0.07364800075689952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,balanced,0.07430399954319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,balanced,0.08123733103275299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,balanced,0.08936533331871033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,balanced,0.09698133667310078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,balanced,0.12377599875132243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,balanced,0.13829867045084634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,balanced,0.18265599012374878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,balanced,0.21715732415517172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,balanced,0.3142559925715129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,balanced,0.4070880015691121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,power_law_1.2,0.30880639553070066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,power_law_1.2,0.4333951950073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,power_law_1.2,0.5540736198425293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,balanced,0.050704002380371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,balanced,0.060975998640060425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,balanced,0.08534933129946391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,balanced,0.11909866333007812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,balanced,0.13447999954223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,balanced,0.13589866956075033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,balanced,0.1371893286705017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,balanced,0.1377066671848297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,balanced,0.13763200243314108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,balanced,0.1397546629110972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,balanced,0.14087466398874918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,balanced,0.1400266687075297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,balanced,0.14416533708572388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,balanced,0.14756799737612405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,balanced,0.15049599607785544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,balanced,0.1535360018412272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,balanced,0.16407466928164163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.17388800779978433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.204693337281545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.20427199204762778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.2632746696472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.284661332766215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.39311468601226807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.4499146540959676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.6544320185979208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,balanced,0.7852906386057535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.05248640179634094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.05354239940643311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.10255999565124511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.14869760274887084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.17906559705734254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.2564800024032593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.41425280570983886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.5576320171356202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,balanced,0.0757066657145818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,balanced,0.10890666643778484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,balanced,0.1696000099182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,balanced,0.2860586643218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,balanced,0.2884906729062398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,balanced,0.2945280075073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,balanced,0.2971786657969157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,balanced,0.296944002310435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,balanced,0.30291734139124554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,balanced,0.31037867069244385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,balanced,0.31170133749643963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,balanced,0.3157493273417155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,balanced,0.32580800851186115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,balanced,0.3311626712481181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,balanced,0.3463946580886841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,balanced,0.38818665345509845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,balanced,0.42056532700856525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,balanced,0.49292266368865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,balanced,0.5498453378677368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,balanced,0.7465386390686035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,balanced,0.8445759614308676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,balanced,1.2126133441925049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,balanced,1.4773279825846355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,balanced,2.1241706212361655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,balanced,2.7403732935587564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,power_law_1.2,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,power_law_1.2,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,power_law_1.2,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,power_law_1.2,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,power_law_1.2,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,power_law_1.2,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,power_law_1.2,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,power_law_1.2,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,power_law_1.2,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,power_law_1.2,0.11004159450531006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,power_law_1.2,0.13678079843521118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,power_law_1.2,0.2107840061187744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,power_law_1.2,0.25579519271850587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,power_law_1.01,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,power_law_1.01,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,power_law_1.01,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,power_law_1.01,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,power_law_1.01,0.061433601379394534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,power_law_1.01,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,power_law_1.01,0.07822080254554749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,power_law_1.01,0.07967360019683838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,power_law_1.01,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,power_law_1.01,0.12963839769363403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,power_law_1.01,0.15430400371551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,power_law_1.01,0.1963263988494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,power_law_1.01,0.25982720851898194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,power_law_1.01,0.29786880016326905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,power_law_1.01,0.4508863925933838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,power_law_1.01,0.5876224040985107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,power_law_1.2,0.37221760749816896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,power_law_1.2,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,power_law_1.2,0.11092480421066284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,power_law_1.2,0.1525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,power_law_1.2,0.2464063882827759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,power_law_1.2,0.36302719116210935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,power_law_1.2,0.5019584178924561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,power_law_1.2,0.5907008171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,power_law_1.2,0.7920256137847901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,power_law_1.2,0.865171241760254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,power_law_1.2,0.8640383720397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,power_law_1.2,0.9165568351745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,power_law_1.2,0.9550399780273438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,power_law_1.2,0.9812992095947266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,power_law_1.2,1.0366144180297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,power_law_1.2,1.0803584098815917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,power_law_1.2,1.0828096389770507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,power_law_1.2,1.1769087791442872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,power_law_1.2,1.2609087944030761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,power_law_1.2,1.3255231857299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,power_law_1.2,1.4849535942077636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,power_law_1.2,1.5635007858276366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,power_law_1.2,1.888697624206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,power_law_1.2,2.0332992553710936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,power_law_1.2,2.5112512588500975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,power_law_1.2,3.0816959381103515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,power_law_1.2,3.962502288818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,power_law_1.2,0.4301248073577881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,power_law_1.2,4.9144126892089846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.0917952001094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.1548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.1915071964263916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.2730047941207886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.3942591905593872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.49748477935791013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.6396096229553223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.2,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.2,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.2,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.2,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.2,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.2,0.11072640419006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.2,0.13663359880447387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.2,0.15614080429077148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.2,0.2094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.2,0.23223679065704345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.2,0.2981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.2,0.38394238948822024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.2,0.6105088233947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.2,0.6892672061920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.09334400296211243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.11248639822006226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.2103872060775757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.27149438858032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.31332480907440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.09534720182418824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.0932864010334015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.09130880236625671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.09984639883041382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.10435839891433715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.10921599864959716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.1260607957839966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.13636480569839476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.16568319797515868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.1909183979034424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.2414016008377075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.27325439453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.3778752088546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.4697472095489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.622867202758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,0.8694016456604003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.1047935962677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.11515519618988038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.129094398021698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.1276352047920227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.1473536014556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.1397696018218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.14457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.13679360151290892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.14706560373306274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.15143040418624878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.15239039659500123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.1700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.19043840169906617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.1990847945213318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.23840639591217042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.2742975950241089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.364300799369812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.40019841194152833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.5102335929870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.6361472129821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,0.8995136260986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.2353280067443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,power_law_1.01,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,power_law_1.01,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,power_law_1.01,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,power_law_1.01,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,power_law_1.01,0.07306240200996399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,power_law_1.01,0.07786239981651306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,power_law_1.01,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,power_law_1.01,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,power_law_1.01,0.0864512026309967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,power_law_1.01,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,power_law_1.01,0.09809920191764832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,power_law_1.01,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,power_law_1.01,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,power_law_1.01,0.14238719940185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,power_law_1.01,0.16908799409866332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,power_law_1.01,0.21303679943084716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.09073920249938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,power_law_1.01,0.2503743886947632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,power_law_1.01,0.34722559452056884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,power_law_1.01,0.43313918113708494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.1041856050491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,power_law_1.01,0.5822336196899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,power_law_1.01,0.7593279838562011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.1146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.10694400072097779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.10762239694595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.11315200328826905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.1130944013595581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.12851840257644653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.12883199453353883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.14008320569992067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.15486719608306884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.16766079664230346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.20461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.2300352096557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.2948607921600342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.331494402885437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.4365695953369141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.5377024173736572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.7224768161773681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.950489616394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,balanced,0.05241066714127859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,balanced,0.04747200012207031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,balanced,0.07755200068155925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,balanced,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,balanced,0.11718400319417317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,balanced,0.11548800269762675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,balanced,0.11717333396275838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,balanced,0.11818666259447734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,balanced,0.11892799536387126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,balanced,0.11972799897193909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,balanced,0.1199626624584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,balanced,0.12144533793131511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,balanced,0.12381333112716675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,balanced,0.12587199608484903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,balanced,0.12614400188128153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,balanced,0.13306132952372232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,balanced,0.1340000033378601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,balanced,0.16555200020472208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,balanced,0.1511626640955607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,balanced,0.21594132979710898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,balanced,0.19924267133076987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,balanced,0.2541439930597941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,balanced,0.28204800685246784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,balanced,0.4469919999440511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,balanced,0.45905065536499023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,balanced,0.08096000055472057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,balanced,0.08505066235860188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,balanced,0.08705066641171773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,balanced,0.08501866459846497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,balanced,0.08736000458399455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,balanced,0.08728532989819844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,balanced,0.08758933345476787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,balanced,0.09006933371225993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,balanced,0.10271466771761577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,balanced,0.10477866729100545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,balanced,0.10474666953086853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,balanced,0.11262399951616923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,balanced,0.12019733587900798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,balanced,0.1362506647904714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,balanced,0.14663466811180115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,balanced,0.18043732643127441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,balanced,0.20173867543538412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,balanced,0.27192533016204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,balanced,0.32173333565394086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,balanced,0.43263999621073407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,balanced,0.5665760040283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.05372160077095032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.14183679819107056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.18190079927444458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.22381439208984374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,power_law_1.01,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,power_law_1.01,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,power_law_1.01,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,power_law_1.01,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,power_law_1.01,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,power_law_1.01,0.058303999900817874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,power_law_1.01,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,power_law_1.01,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,power_law_1.01,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,power_law_1.01,0.13393919467926024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,power_law_1.01,0.18152960538864135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,power_law_1.01,0.21647360324859619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,power_law_1.01,0.29482879638671877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,power_law_1.01,0.388755202293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,power_law_1.01,0.565228796005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,power_law_1.2,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,power_law_1.2,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,power_law_1.01,0.7449408054351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,power_law_1.2,0.052076798677444455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,power_law_1.2,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,power_law_1.2,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,power_law_1.2,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,power_law_1.2,0.12585599422454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,power_law_1.2,0.1665984034538269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,power_law_1.2,0.21160318851470947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,power_law_1.2,0.24282240867614746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,power_law_1.2,0.34815359115600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,power_law_1.2,0.46151041984558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,power_law_1.2,0.6719744205474854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,power_law_1.2,0.8691007614135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.3037568092346191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.38139519691467283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,power_law_1.01,0.5434815883636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,power_law_1.01,0.6832191944122314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,power_law_1.2,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,power_law_1.2,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,power_law_1.2,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,power_law_1.2,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,power_law_1.2,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,power_law_1.2,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,power_law_1.2,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,power_law_1.2,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,power_law_1.2,0.09054719805717468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,power_law_1.2,0.10046720504760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,power_law_1.2,0.10067199468612671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,power_law_1.2,0.10947200059890747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,power_law_1.2,0.12075519561767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.2,0.1289023995399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.2,0.16329599618911744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.2,0.20289919376373292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.2,0.23973119258880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.2,0.2960319995880127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.2,0.4493696212768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.2,0.6208896160125732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.2,0.9159232139587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.2,1.4734272003173827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.01,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.01,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.01,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.01,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.01,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,balanced,0.04825599988301595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,balanced,0.040048000713189445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,balanced,0.04837333162625631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,balanced,0.0492799977461497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,balanced,0.05054933329423269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,balanced,0.05570666491985321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,balanced,0.05524266759554545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,balanced,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,balanced,0.06398400167624156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,balanced,0.07561066746711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,balanced,0.08758399883906047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,balanced,0.11063466469446818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,balanced,0.12142399946848552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,balanced,0.15559466679890951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.01,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.01,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.01,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.01,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.01,0.10825599431991577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,balanced,0.1713119943936666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,balanced,0.23505600293477377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.01,0.13038079738616942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,balanced,0.2828106681505839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.01,0.1833024024963379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.01,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,balanced,0.08204266428947449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,balanced,0.08213866750399272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,balanced,0.08398933211962382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,balanced,0.08561066786448161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,balanced,0.08880533774693807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,balanced,0.08899733424186707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,balanced,0.09064533313115437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,balanced,0.09724799791971843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,balanced,0.09690666198730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,balanced,0.10099200407663982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,balanced,0.11317333579063416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,balanced,0.12593600153923035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,balanced,0.1458186705907186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,balanced,0.17421332995096842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,balanced,0.21224000056584677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,balanced,0.2627626657485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,balanced,0.38280534744262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,balanced,0.4580479860305786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,balanced,0.6605279843012491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,balanced,0.8437440395355225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,balanced,0.06398400167624156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,balanced,0.06377600133419037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,balanced,0.06884799897670746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,balanced,0.07065066695213318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,balanced,0.08029866715272267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.09619733691215515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.10938666264216106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.13751467068990073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.1613599956035614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.21042132377624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2521493236223857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.34700266520182294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.4350080092748006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,balanced,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,balanced,0.047610665361086525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,balanced,0.07308266560236613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,balanced,0.09789333740870158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,balanced,0.12636267145474753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,balanced,0.1262453297773997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,balanced,0.1276639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,balanced,0.12770666678746542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,balanced,0.12652800480524698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,balanced,0.12864533066749573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,balanced,0.12796266873677573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,balanced,0.1309653321901957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,balanced,0.13562132914861044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,balanced,0.13546666502952576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,balanced,0.13929067055384317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,balanced,0.1450399955113729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,balanced,0.15004799763361612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,balanced,0.1602666676044464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,balanced,0.171615997950236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,balanced,0.19400533040364584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,balanced,0.21996800104777017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,balanced,0.25999999046325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,balanced,0.32366933425267536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,balanced,0.4269760052363078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,balanced,0.5048319896062216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,balanced,0.05333333214124044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,balanced,0.06621866424878438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,balanced,0.06621333460013072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,balanced,0.06753066678841908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,balanced,0.06866133213043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,balanced,0.06917333106199901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,balanced,0.07884799937407176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,balanced,0.08290133376916249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,balanced,0.08874666690826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.09483200311660767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.11382933457692464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.13131200273831686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.1622880001862844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.19163199265797934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.25058666865030926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.3126773238182068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.4309813181559245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.55021866162618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.06254079937934875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.08443520069122315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.09269760251045227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.1560639977455139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.18920960426330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.22206718921661378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.31729280948638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.3688640117645264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,power_law_1.01,0.5409855842590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,power_law_1.01,0.7307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.07519999742507935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.1399616003036499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.21940479278564454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.28563199043273924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.3701440095901489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.5689216136932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,power_law_1.2,0.6476672172546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,power_law_1.2,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,power_law_1.2,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,power_law_1.2,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,power_law_1.2,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,power_law_1.2,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,power_law_1.2,0.07804160118103028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,power_law_1.2,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,power_law_1.2,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,power_law_1.2,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,power_law_1.2,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,power_law_1.2,0.1040768027305603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,power_law_1.2,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,power_law_1.2,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,power_law_1.2,0.12737280130386353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,power_law_1.2,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,power_law_1.2,0.1662335991859436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,power_law_1.2,0.21175680160522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,power_law_1.2,0.2644223928451538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,power_law_1.2,0.3292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,power_law_1.2,0.45498242378234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,power_law_1.2,0.5441343784332275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,power_law_1.2,0.7987008094787598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,power_law_1.2,1.082489585876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.2,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.2,0.028620800375938414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.2,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.2,0.030015999078750612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.2,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.2,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.2,0.07086079716682434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.2,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.2,0.10782719850540161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.2,0.1310912013053894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.2,0.23071999549865724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,power_law_1.2,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.2,0.2260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,power_law_1.2,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,power_law_1.2,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,power_law_1.2,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,power_law_1.2,0.056460797786712646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,power_law_1.2,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,power_law_1.2,0.05963519811630249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,power_law_1.2,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,power_law_1.2,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,power_law_1.2,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,power_law_1.2,0.06908800005912781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,power_law_1.2,0.07446399927139283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,power_law_1.2,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.2,0.09452800154685974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.2,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.2,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.2,0.1662783980369568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.2,0.20174078941345214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.2,0.28225278854370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.2,0.36389119625091554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.2,0.5463168144226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.2,0.7578879833221436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.08743039965629577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.08296319842338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.08620799779891967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.09384959936141968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.09326720237731934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.106387197971344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.11507840156555176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.15073280334472655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.17973120212554933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.20892159938812255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.27365119457244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.3537535905838013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.486297607421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,power_law_1.01,0.5931903839111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.07205119729042053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.07478399872779846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.08659840226173401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.12748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.14714239835739135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.19007359743118285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.22204160690307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.30065920352935793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.37496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.5228096008300781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.705836820602417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.05601919889450073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.08724480271339416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.11235200166702271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.12936320304870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.14321279525756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.17656320333480835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.22707839012145997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.27352960109710694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.40018558502197266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.4328127861022949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.7339647769927978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,0.8052736282348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,power_law_1.01,0.08616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,power_law_1.01,0.09000959992408752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,power_law_1.01,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,power_law_1.01,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,power_law_1.01,0.09158400297164918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,power_law_1.01,0.09276800155639649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,power_law_1.01,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,power_law_1.01,0.09555839896202087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,power_law_1.01,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,power_law_1.01,0.11480319499969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,power_law_1.01,0.13631999492645264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,power_law_1.01,0.1530303955078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,power_law_1.01,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,power_law_1.01,0.17432960271835327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,power_law_1.01,0.27513599395751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.01,0.32697598934173583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.01,0.43841280937194826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.01,0.5575744152069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.01,0.7225215911865235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.01,1.061235237121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.01,1.5707584381103517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.01,1.8606271743774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,power_law_1.01,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,power_law_1.01,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,power_law_1.01,0.09970560073852539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,power_law_1.01,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,power_law_1.01,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,power_law_1.01,0.10716160535812377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,power_law_1.01,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,power_law_1.01,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,power_law_1.01,0.11685760021209717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,power_law_1.01,0.11738879680633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,power_law_1.01,0.12158080339431762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,power_law_1.01,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,power_law_1.01,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,power_law_1.01,0.1438591957092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,power_law_1.01,0.1577023983001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.01,0.17393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.01,0.20622720718383789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.01,3.1628095626831056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.01,0.24895360469818115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.01,0.31313281059265136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.01,0.37563519477844237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.01,0.5349952220916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.01,0.6785088062286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.01,0.9614527702331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.01,1.2539967536926269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.01,4.130342483520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,power_law_1.01,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,power_law_1.01,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,power_law_1.01,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,power_law_1.01,0.055404800176620486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,power_law_1.01,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,power_law_1.01,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,power_law_1.01,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,power_law_1.01,0.08535680174827576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,power_law_1.01,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,power_law_1.01,0.11935360431671142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,power_law_1.01,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,power_law_1.01,0.1586624026298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,power_law_1.01,0.19633920192718507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,power_law_1.01,0.2834815979003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,power_law_1.01,0.393286395072937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,power_law_1.01,0.4682943820953369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,power_law_1.01,0.6982912063598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.12162560224533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.12171519994735717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.1235967993736267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.12214399576187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.12212480306625366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.1546496033668518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.19526400566101074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.21543679237365723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.24723200798034667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.15390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.16533119678497316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.1693824052810669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.18837120532989501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.24506878852844238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.23539841175079346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.2637887954711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,power_law_1.2,0.3478656053543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,power_law_1.2,0.44305920600891113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,power_law_1.2,0.569811201095581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,power_law_1.2,0.6301695823669433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,power_law_1.2,0.9887935638427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,power_law_1.2,1.2722944259643554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,power_law_1.2,1.9611648559570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,power_law_1.2,2.301350402832031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.01,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.01,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.01,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.01,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.01,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.01,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.01,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.01,0.082777601480484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.01,0.11766400337219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.01,0.13521920442581176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.01,0.1801344037055969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.01,0.18799359798431398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.01,0.24769918918609618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.01,0.28884479999542234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.01,0.3977855920791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.01,0.4810368061065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,power_law_1.2,3.873779296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,power_law_1.2,4.487756729125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,balanced,0.05060799916585287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,balanced,0.0653653343518575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,balanced,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,balanced,0.128602663675944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,balanced,0.19957866271336874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,balanced,0.2549813389778137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,balanced,0.25600000222524005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,balanced,0.26073600848515827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,balanced,0.2584213415781657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,balanced,0.2602826754252116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,balanced,0.2619040012359619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,balanced,0.265120009581248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,balanced,0.26707732677459717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,balanced,0.2696320017178853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,balanced,0.27877867221832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,balanced,0.2823573350906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,balanced,0.29043734073638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,balanced,0.34759998321533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,balanced,0.3282453417778015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,balanced,0.4198773304621379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,balanced,0.41701332728068036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,balanced,0.5823520024617513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,balanced,0.5715573231379191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,balanced,0.7818079789479574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,balanced,0.8807306289672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,balanced,1.3149653275807698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,balanced,1.5302666028340657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.05601919889450073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.08207359910011292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.08473600149154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.09204480051994324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.10257279872894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.1224128007888794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.13544319868087767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.184607994556427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.2315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.28512001037597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.4031807899475098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.5399168014526368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.6801152229309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,0.9391360282897949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.2998592376708984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,balanced,0.048485333720842995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,balanced,0.06833600004514058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,balanced,0.09416533509890239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,balanced,0.14756266276041666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,balanced,0.24658666054407755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,balanced,0.4429440100987752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,balanced,0.44546135266621906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,balanced,0.4464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,balanced,0.44973333676656085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,balanced,0.44844265778859455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,balanced,0.4548319975535075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,balanced,0.45923733711242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,balanced,0.4604533513387044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,balanced,0.47148799896240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,balanced,0.48348267873128253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,balanced,0.48906668027242023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,balanced,0.4979413350423177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,balanced,0.5227146546045939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,balanced,0.5440959930419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,balanced,0.602570652961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,balanced,0.6559520165125529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,balanced,0.746293306350708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,balanced,0.8432693481445312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,balanced,1.112661361694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,balanced,1.265770673751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,balanced,1.8000853856404622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,balanced,2.232288042704264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,balanced,0.03595733394225439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,balanced,0.06923733154932658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,balanced,0.1046346624692281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,balanced,0.17057067155838013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,balanced,0.21851734320322672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,balanced,0.22028799851735434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,balanced,0.22249066829681396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,balanced,0.22338666518529257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,balanced,0.22557334105173746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,balanced,0.22563733657201132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,balanced,0.22853867212931314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,balanced,0.23453332980473837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,balanced,0.23089067141215006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,balanced,0.23907200495402017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,balanced,0.2610666751861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,balanced,0.24409600098927817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,balanced,0.29022934039433795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,balanced,0.26844799518585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,balanced,0.3185386657714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,balanced,0.3084213336308797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,balanced,0.42743468284606934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,balanced,0.41489601135253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,balanced,0.5881866614023844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,balanced,0.6535733143488566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,balanced,0.9514400164286295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,balanced,1.1026559670766194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.15018880367279053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.2399616003036499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.2798847913742065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.32466559410095214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.40305280685424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.3676352024078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.3960832118988037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.40003199577331544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.3647295951843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.3799488067626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.3887936115264893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.3874495983123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.42052478790283204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.41649279594421384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.40782718658447265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.4294144153594971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.45608320236206057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.4736447811126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.5017983913421631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.5834112167358398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,power_law_1.01,0.5521855831146241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,power_law_1.01,0.645689582824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,power_law_1.01,0.7211647987365722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,power_law_1.01,0.8933183670043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,power_law_1.01,1.0640064239501954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,power_law_1.01,1.4475392341613769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,power_law_1.01,1.7425535202026368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,power_law_1.2,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,power_law_1.2,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,power_law_1.2,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,power_law_1.2,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,power_law_1.2,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,power_law_1.2,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,power_law_1.2,0.06751999855041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,power_law_1.2,0.07007359862327575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,power_law_1.2,0.07107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,power_law_1.2,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,power_law_1.2,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,power_law_1.2,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,power_law_1.2,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.2,0.13652479648590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.2,0.18723200559616088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.2,0.23527040481567382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.2,0.34339199066162107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.2,0.4210048198699951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.2,0.5689280033111572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.2,0.8370944023132324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,power_law_1.01,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,power_law_1.01,0.11152000427246093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,power_law_1.01,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,power_law_1.01,0.2018496036529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,power_law_1.01,0.2663615942001343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,power_law_1.01,0.32471680641174316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,power_law_1.01,0.4156479835510254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,power_law_1.01,0.46737918853759763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,power_law_1.01,0.4591807842254639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,power_law_1.01,0.47002239227294923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,power_law_1.01,0.47895040512084963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,power_law_1.01,0.48586239814758303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,power_law_1.01,0.4809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,power_law_1.01,0.48389759063720705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,power_law_1.01,0.5045248031616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,power_law_1.01,0.5360191822052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,power_law_1.01,0.5543039798736572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,power_law_1.01,0.5961023807525635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,power_law_1.01,0.5912000179290772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,power_law_1.01,0.6592959880828857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,power_law_1.01,0.6942399978637696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,power_law_1.01,0.8164799690246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,power_law_1.01,0.8504639625549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,power_law_1.01,1.0726207733154296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,power_law_1.01,1.1889472007751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.2,1.3307519912719727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,power_law_1.01,1.56943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.2,1.7113023757934571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,power_law_1.01,1.9087104797363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,power_law_1.01,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,power_law_1.01,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,power_law_1.01,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,power_law_1.01,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,power_law_1.01,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.01,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.01,0.16647039651870726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,balanced,0.05691733459631602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,balanced,0.06123200058937073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,balanced,0.09363733728726704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,balanced,0.12516799569129944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,balanced,0.12657066186269125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,balanced,0.12964800000190735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,balanced,0.12742933630943298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,balanced,0.12966400384902954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,balanced,0.1290720005830129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.01,0.18255360126495362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,balanced,0.13000532984733582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,balanced,0.13173333803812662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,balanced,0.1339040001233419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,balanced,0.1425973375638326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,balanced,0.14594133694966635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,balanced,0.14877866705258688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,balanced,0.15902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,balanced,0.16541866461435953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,balanced,0.1932106614112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,balanced,0.2151306668917338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,balanced,0.2583093245824178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,balanced,0.29662400484085083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,balanced,0.41275731722513836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,balanced,0.5006453196207682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,balanced,0.706490675608317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.01,0.24293758869171142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,balanced,0.9176212946573893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.01,0.34406399726867676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.01,0.5660352230072021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.01,0.744761610031128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.01,1.1150143623352051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.01,1.3969087600708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.11514240503311157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.15742080211639403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.2060352087020874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,power_law_1.2,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,power_law_1.2,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,power_law_1.2,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,power_law_1.2,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,power_law_1.2,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,power_law_1.2,0.08795520067214965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,power_law_1.2,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,power_law_1.2,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,power_law_1.2,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,power_law_1.2,0.0902783989906311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,power_law_1.2,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,power_law_1.2,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,power_law_1.2,0.10138239860534667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,power_law_1.2,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,power_law_1.2,0.1305343985557556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,power_law_1.2,0.1392575979232788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,power_law_1.2,0.1720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,power_law_1.2,0.19842560291290284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,power_law_1.2,0.25128960609436035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,power_law_1.2,0.3274431943893433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,power_law_1.2,0.44824957847595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,power_law_1.2,0.6303423881530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,power_law_1.2,0.7940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,power_law_1.2,1.0874239921569824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,balanced,0.03992533435424169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,balanced,0.09052800138791402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,balanced,0.09091200431187947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,balanced,0.09276800354321797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,balanced,0.09179199735323589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,balanced,0.09207466244697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,balanced,0.09363200267155965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,balanced,0.09384000301361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,balanced,0.09345066547393799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,balanced,0.10148800412813823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,balanced,0.10082667072614034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,balanced,0.10681600371996562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,balanced,0.11358933647473653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,balanced,0.12176533540089925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,balanced,0.13573333621025085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,balanced,0.14841600259145102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,balanced,0.17680533727010092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,balanced,0.21450666586558023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,balanced,0.26740799347559613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,balanced,0.32443199555079144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,balanced,0.4396053155263265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,balanced,0.5514933268229166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,power_law_1.01,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,power_law_1.01,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,power_law_1.01,0.10382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,power_law_1.01,0.11077760457992554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,power_law_1.01,0.11488640308380127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,power_law_1.01,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,power_law_1.01,0.11959680318832397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,power_law_1.01,0.1214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,power_law_1.01,0.12208000421524048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,power_law_1.01,0.12611839771270753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,power_law_1.01,0.12637439966201783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,power_law_1.01,0.13860479593276978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,power_law_1.01,0.1423424005508423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,power_law_1.01,0.15140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,power_law_1.01,0.17904000282287597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,power_law_1.01,0.18485759496688842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,power_law_1.01,0.2288896083831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,power_law_1.01,0.26917760372161864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,power_law_1.01,0.3421823978424072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,power_law_1.01,0.433241605758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,power_law_1.01,0.555788803100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,power_law_1.01,0.7680831909179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,power_law_1.01,0.9702336311340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,power_law_1.01,1.3598976135253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,power_law_1.01,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,power_law_1.01,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,power_law_1.01,0.11098239421844483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,power_law_1.01,0.12144639492034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,power_law_1.01,0.13407360315322875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,power_law_1.01,0.1486207962036133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,power_law_1.01,0.15263359546661376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,power_law_1.01,0.1578752040863037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,power_law_1.01,0.15838079452514647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,2,power_law_1.01,0.16341760158538818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,2,power_law_1.01,0.16288000345230103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,2,power_law_1.01,0.16844160556793214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,2,power_law_1.01,0.17002880573272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,2,power_law_1.01,0.17430399656295775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,2,power_law_1.01,0.17668479681015015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,2,power_law_1.01,0.18492159843444825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,2,power_law_1.01,0.20262401103973388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.01,0.21685760021209716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.01,0.25528318881988527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.01,0.29427199363708495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.01,0.3635456085205078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.01,0.4575488090515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.01,0.6195136070251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.01,0.8143424034118653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.01,1.1239680290222167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.01,1.4385791778564454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,power_law_1.01,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,power_law_1.01,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,power_law_1.01,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,power_law_1.01,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,power_law_1.01,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,power_law_1.01,0.0822655975818634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,power_law_1.01,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,power_law_1.01,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.01,0.11203839778900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.01,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.01,0.18016639947891236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.01,0.22529280185699463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.01,0.3019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.01,0.39813120365142823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.01,0.6162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.01,0.7727615833282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,power_law_1.01,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,power_law_1.01,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,power_law_1.01,0.06713600158691406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,power_law_1.01,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,power_law_1.01,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,power_law_1.01,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,power_law_1.01,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,power_law_1.01,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,power_law_1.01,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,power_law_1.01,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,power_law_1.01,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,power_law_1.01,0.08475520014762879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,power_law_1.01,0.08734080195426941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,power_law_1.01,0.08968960046768189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,power_law_1.01,0.0917952001094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,power_law_1.01,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.01,0.10817919969558716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.01,0.13318400382995604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.01,0.1475775957107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.01,0.18471039533615113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.01,0.21727359294891357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.01,0.3153791904449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.01,0.40838398933410647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.01,0.6250495910644531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.01,0.7137599945068359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.2,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.2,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.2,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.2,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.2,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.2,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.2,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.2,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.2,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.2,0.11036800146102906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.2,0.14985599517822265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.2,0.21497600078582763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.2,0.2927808046340942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.2,0.3126271963119507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,balanced,0.05133866767088572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,balanced,0.0620000014702479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,balanced,0.0883146623770396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,balanced,0.1227946678797404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,balanced,0.16421332955360413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,balanced,0.19751467307408652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,balanced,0.2657066583633423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,balanced,0.33714667956034344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,balanced,0.07573866844177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,balanced,0.10132799545923869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,balanced,0.15332800149917603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,balanced,0.2551199992497762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,balanced,0.4657813310623169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,balanced,0.5675626595815023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,balanced,0.5719626744588217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,balanced,0.5731626749038696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,balanced,0.5756693283716837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,balanced,0.575984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,balanced,0.5795573393503824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,balanced,0.5848106543223063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,balanced,0.5859040021896362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,balanced,0.5903573433558146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,balanced,0.6017280022303263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,balanced,0.6042986710866293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,balanced,0.6151146491368612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,balanced,0.6572959820429484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,balanced,0.6528426806131998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,balanced,0.77020796140035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,balanced,0.7375840346018473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,balanced,1.140336036682129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,balanced,0.9031306902567545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,balanced,1.3046987056732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,balanced,1.3570613861083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,balanced,2.1812853813171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,balanced,2.3911733627319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.01,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.01,0.18763519525527955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.01,0.16960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.01,0.19748480319976808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.01,0.22693119049072266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.01,0.22060160636901854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.01,0.2349951982498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.01,0.23034238815307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.01,0.21640961170196532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.01,0.22902400493621827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.01,0.23934719562530518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.01,0.23720960617065429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.01,0.2465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.01,0.24976000785827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.01,0.2610111951828003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.01,0.26720640659332273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.01,0.2837503910064697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.01,0.31626880168914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.01,0.34743680953979494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.01,0.4254335880279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.01,0.4553215980529785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.01,0.5505983829498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.01,0.614028787612915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.01,0.7570752143859864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.01,0.8924736022949219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.01,1.1816320419311523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.01,1.470195198059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,balanced,0.047024001677831016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,balanced,0.04615999758243561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,balanced,0.04753066599369049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,balanced,0.049141332507133484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,balanced,0.05643733342488607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,balanced,0.05644799768924713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,balanced,0.07826133569081624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,balanced,0.08728532989819844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,balanced,0.11763200163841248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,balanced,0.13821333646774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,power_law_1.2,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,balanced,0.19026132424672446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,balanced,0.24505066871643066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,power_law_1.2,0.022265599668025972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,balanced,0.3476959864298503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,balanced,0.44121599197387695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,power_law_1.2,0.022060799598693847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,power_law_1.2,0.02333440035581589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,power_law_1.2,0.022060799598693847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,power_law_1.2,0.024582399427890776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,power_law_1.2,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,power_law_1.2,0.027001601457595826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,power_law_1.2,0.028224000334739686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,power_law_1.2,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,power_law_1.2,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,power_law_1.2,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,power_law_1.2,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,power_law_1.2,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,power_law_1.2,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.08564479947090149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.11371519565582275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,power_law_1.2,0.06661120057106018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,power_law_1.2,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,power_law_1.2,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.18337279558181763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,power_law_1.2,0.07011200189590454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,power_law_1.2,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,power_law_1.2,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,power_law_1.2,0.07619839906692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,power_law_1.2,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,power_law_1.2,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,power_law_1.2,0.09160320162773132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,power_law_1.2,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,power_law_1.2,0.09884160161018371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,power_law_1.2,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,power_law_1.2,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,power_law_1.2,0.14858880043029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,power_law_1.2,0.17172479629516602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,power_law_1.2,0.24012160301208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,power_law_1.2,0.2677119970321655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,power_law_1.2,0.3925120115280151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,power_law_1.2,0.48959999084472655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,power_law_1.2,0.8012607574462891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,power_law_1.2,0.9302271842956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,balanced,0.03396799912055334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,balanced,0.035962666074434914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,balanced,0.04417066772778829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,balanced,0.05346133311589559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,balanced,0.05574400226275126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,balanced,0.058890665570894875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,balanced,0.07300266623497009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,balanced,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,balanced,0.1172213355700175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,balanced,0.15252799789110819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,balanced,0.18359466393788657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,balanced,0.26023467381795246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,balanced,0.3407520055770874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,power_law_1.2,0.04871039986610413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,power_law_1.2,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,power_law_1.2,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,power_law_1.2,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,power_law_1.2,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,power_law_1.2,0.05928959846496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,power_law_1.2,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,power_law_1.2,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,power_law_1.2,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,power_law_1.2,0.11477760076522828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,power_law_1.2,0.1540992021560669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,power_law_1.2,0.18979840278625487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,power_law_1.2,0.2464128017425537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,power_law_1.2,0.31992321014404296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,power_law_1.2,0.4561471939086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,power_law_1.2,0.6126976013183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.1273408055305481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.15752320289611815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.21954560279846191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.2319488048553467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.23955841064453126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.24134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.25426559448242186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.2632191896438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.2661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.27480320930480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.2828736066818237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.29443840980529784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.30293118953704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.3199680089950562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.3704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.4097343921661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.5054143905639649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.5791423797607422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.7594175815582276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.9289407730102539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.2641535758972169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,1.7472320556640626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.3975807189941407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,2.99051513671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.09658880233764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.09697920083999634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.12155519723892212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.12881920337677003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.11139199733734131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.12624000310897826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.12930560111999512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.12465280294418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.1256767988204956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.13596160411834718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.1456768035888672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.16054400205612182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.17430399656295775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.20592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.21409280300140382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.26439039707183837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.3267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.4198783874511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,power_law_1.01,0.5527872085571289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,power_law_1.01,0.7734335899353028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,power_law_1.01,0.8807999610900878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.028095999360084535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.028384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.1102720022201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.14574719667434693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.18595199584960936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.24868481159210204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.29656319618225097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.3452415943145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,balanced,0.051226665576299034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,balanced,0.05166399975617727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,balanced,0.05065600077311198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,balanced,0.05576533575852712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,balanced,0.06449066599210103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,balanced,0.06647466619809468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,balanced,0.07259733478228252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,balanced,0.0724533349275589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,balanced,0.07452266911665599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,balanced,0.08275733391443889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.1011306643486023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.12359999616940816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.14824533462524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.17123200496037802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.23031467199325562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.27897600332895917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.3879679838816325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.49272533257802326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,power_law_1.01,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,power_law_1.01,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,power_law_1.01,0.12832000255584716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,power_law_1.01,0.163756799697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,power_law_1.01,0.18124799728393554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,power_law_1.01,0.21422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,power_law_1.01,0.22540159225463868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,power_law_1.01,0.231660795211792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,power_law_1.01,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,power_law_1.01,0.25169920921325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,power_law_1.01,0.26289920806884765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,power_law_1.01,0.2811072111129761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,power_law_1.01,0.3006335973739624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,power_law_1.01,0.3198784112930298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,power_law_1.01,0.3187328100204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,power_law_1.01,0.3720576047897339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,power_law_1.01,0.41170558929443357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,power_law_1.01,0.5115007877349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,power_law_1.01,0.5966207981109619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,power_law_1.01,0.7783872127532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,power_law_1.01,0.9645631790161133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,power_law_1.01,1.3061375617980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,power_law_1.01,1.621126365661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,power_law_1.01,2.292915153503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,power_law_1.01,2.983500862121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,balanced,0.05219733218352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,balanced,0.13641599814097086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,balanced,0.1814133326212565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,balanced,0.16765866676966348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,balanced,0.16820800304412842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,balanced,0.1695680022239685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,balanced,0.17056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,balanced,0.1759679913520813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,balanced,0.17780266205469766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,balanced,0.17808000246683756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,balanced,0.18920000394185385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,balanced,0.19236266613006592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,balanced,0.202949325243632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,balanced,0.22395733992258707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,balanced,0.23997867107391357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,balanced,0.28327999512354535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,balanced,0.3189653356870015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,balanced,0.40353600184122723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,balanced,0.5017600059509277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,balanced,0.6880373160044352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,balanced,0.8575680255889893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,balanced,1.2318027019500732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,balanced,1.6098027229309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06346240043640136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.07619839906692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.11822719573974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.13899519443511962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.17786879539489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.2094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.2974463939666748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.37994239330291746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.01,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.01,0.0605567991733551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.01,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.01,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.01,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.551475191116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.01,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.01,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.01,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.01,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.01,0.10053759813308716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.01,0.11373440027236939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.01,0.13963520526885986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.01,0.1689792037010193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.01,0.22077438831329346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.01,0.23281919956207275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.01,0.29914240837097167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.01,0.3785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.01,0.5252223968505859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.01,0.6310080051422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.6927167892456054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,1.0233920097351075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,balanced,0.04347200194994608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,balanced,0.04761599997679392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,balanced,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,balanced,0.06774400174617767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,balanced,0.06678933401902516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,balanced,0.06861333549022675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,balanced,0.06883733471234639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,balanced,0.06877333422501881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,balanced,0.07246399919191997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,balanced,0.07727999985218048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,balanced,0.07930666704972585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,balanced,0.08583999673525493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,balanced,0.08983467022577922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,balanced,0.11750933527946472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,balanced,0.13987732927004495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,balanced,0.17278399070103964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.3259136199951171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,balanced,0.22735466559727988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,balanced,0.2621386647224426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,balanced,0.3644426663716634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,balanced,0.45872000853220624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,balanced,0.03395200024048487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,balanced,0.043493335445721946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,balanced,0.08615466952323914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,balanced,0.10948800047238667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,balanced,0.12060800194740295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,balanced,0.16131200393040976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,balanced,0.18869332472483316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.06949759721755981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.07395840287208558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.08743039965629577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.10172159671783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.13855359554290772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.16219520568847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.20037760734558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.24081919193267823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.33289599418640137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.4559296131134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.5926271915435791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.8365823745727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,power_law_1.2,0.02282239943742752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,power_law_1.2,0.021593600511550903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,power_law_1.2,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,power_law_1.2,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,power_law_1.2,0.022899200022220612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,power_law_1.2,0.02401279956102371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,power_law_1.2,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,power_law_1.2,0.028230398893356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,power_law_1.2,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,power_law_1.2,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,power_law_1.2,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.11295360326766968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.12085119485855103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.1194111943244934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.11990400552749633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.131769597530365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.14305919408798218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.13918720483779906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.15104639530181885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.16637439727783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.1863487958908081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.21560320854187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.24967679977416993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.3259455919265747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.38410239219665526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.5500736236572266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.6767295837402344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,0.9923007965087891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,1.298854446411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.11477760076522828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.13115520477294923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.16641279458999633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.2447808027267456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.2,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.2,0.031891199946403506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.2,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.2,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.2,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.2,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.2,0.10714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.2,0.12709120512008668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.2,0.17170560359954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.2,0.1936959981918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.2,0.29317760467529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.2,0.37740800380706785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,balanced,0.06717866659164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,balanced,0.07063999772071838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,balanced,0.07008533179759979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,balanced,0.06931733091672261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,balanced,0.07063999772071838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,balanced,0.07030400137106578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,balanced,0.07153066496054332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,balanced,0.07283199826876323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,balanced,0.07574399809042613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,balanced,0.07746133208274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,balanced,0.0869813362757365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,balanced,0.09144000212351482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,balanced,0.1013813316822052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,balanced,0.11396267016728719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,balanced,0.1439626713593801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,balanced,0.16086933016777039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,balanced,0.22326399882634482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,balanced,0.28600533803304035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,balanced,0.4203466574350993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,balanced,0.5517439842224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,balanced,0.09301333626111348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,balanced,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,balanced,0.24463466803232828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,balanced,0.44033066431681317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,balanced,0.6381440162658691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,balanced,0.6381066640218099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,balanced,0.6401493151982626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,balanced,0.6405760049819946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,balanced,0.6462933222452799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,balanced,0.6507786512374878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,balanced,0.6521546840667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,balanced,0.6563146511713663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,balanced,0.6642826795578003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,balanced,0.6667253176371256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,balanced,0.679914633433024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,balanced,0.6985440254211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,balanced,0.7147520383199056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,balanced,0.7531786759694418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,balanced,0.7901866436004639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,balanced,0.8587946891784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,balanced,0.9361546834309896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,balanced,1.0698453585306804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,balanced,1.252559979756673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,balanced,1.5107839902242024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,balanced,1.9839785893758137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.08676480054855347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.08744959831237793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.11296639442443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.11727999448776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.17961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.22296319007873536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.26944000720977784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.36199040412902833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.49562878608703614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,power_law_1.2,0.613324785232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,power_law_1.2,0.8686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,power_law_1.2,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,power_law_1.2,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,power_law_1.2,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,power_law_1.2,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,power_law_1.2,0.07839999794960022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,power_law_1.2,0.08011519908905029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,power_law_1.2,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,power_law_1.2,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,power_law_1.2,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,power_law_1.2,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,power_law_1.2,0.100108802318573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,power_law_1.2,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,power_law_1.2,0.12698880434036255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,power_law_1.2,0.1369279980659485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,power_law_1.2,0.162009596824646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,power_law_1.2,1.0760383605957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,power_law_1.2,0.2076416015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,power_law_1.2,0.25295999050140383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,power_law_1.2,0.3087296009063721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,power_law_1.2,0.4516160011291504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,power_law_1.2,0.576038408279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,power_law_1.2,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,power_law_1.2,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,power_law_1.2,0.060236799716949466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,power_law_1.2,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,power_law_1.2,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,power_law_1.2,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,power_law_1.2,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,power_law_1.2,0.12699520587921143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,power_law_1.2,0.1375615954399109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,power_law_1.2,0.16643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,power_law_1.2,0.21709439754486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,power_law_1.2,0.7930047988891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,power_law_1.2,0.3005824089050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,power_law_1.2,0.3486783981323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,power_law_1.2,0.4997824192047119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,power_law_1.2,0.6961088180541992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,power_law_1.2,1.0230912208557128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.04963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.09987840056419373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.12495360374450684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.13894399404525756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.19045759439468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.22203519344329833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.26496639251708987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.32607359886169435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,balanced,0.04645333190759023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,balanced,0.04659200211366018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,balanced,0.051274667183558144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,balanced,0.05845866600672404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,balanced,0.07569600145022075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,balanced,0.07735466460386912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,balanced,0.07785599927107494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,balanced,0.08229866623878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,balanced,0.0842186709245046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,balanced,0.08711999654769897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,balanced,0.08870399991671245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.029183998703956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,balanced,0.0935093363126119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,balanced,0.107232004404068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.5493375778198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,balanced,0.12055466572443645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,balanced,0.14415466785430908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,balanced,0.1669600009918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,balanced,0.19272534052530924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,balanced,0.2534186641375224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,balanced,0.2983253399531047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,balanced,0.41420265038808185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,balanced,0.520138661066691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.08259199857711792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.16070400476455687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.12965760231018067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.17432960271835327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.1807935953140259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.2128063917160034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.24682240486145018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.6801599979400634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.4223360061645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.2807807922363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,power_law_1.01,0.3684096097946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.2,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.2,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.2,0.18065279722213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.2,0.18019839525222778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.2,0.18799999952316285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.2,0.1869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.2,0.19487359523773193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.2,0.2009023904800415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.2,0.19843839406967162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.2,0.20456318855285643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.2,0.20983679294586183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.2,0.21801600456237794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.2,0.22651519775390624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.2,0.244051194190979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.2,0.2721280097961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.2,0.2626240015029907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.2,0.3095232009887695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.2,0.35747199058532714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.2,0.4050303936004639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.2,0.44234237670898435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.2,0.5244800090789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.2,0.6600128173828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.2,0.8557056427001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.2,1.0496767997741698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.048607999086380006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,power_law_1.01,0.04630399942398071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,power_law_1.01,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,power_law_1.01,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.1480064034461975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,power_law_1.01,0.07610880136489868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,power_law_1.01,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,power_law_1.01,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,power_law_1.01,0.10327039957046509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,power_law_1.01,0.12147200107574463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,power_law_1.01,0.14404480457305907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,power_law_1.01,0.1812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,power_law_1.01,0.2300352096557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,power_law_1.01,0.29261438846588134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,power_law_1.01,0.39587841033935545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,power_law_1.01,0.5184703826904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.22445440292358398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.22544639110565184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.4001791954040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,1,power_law_1.01,0.12459520101547242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,1,power_law_1.01,0.14450559616088868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,1,power_law_1.01,0.17148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,1,power_law_1.01,0.2022655963897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,1,power_law_1.01,0.24126079082489013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,1,power_law_1.01,0.2753024101257324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,1,power_law_1.01,0.3366976022720337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,1,power_law_1.01,0.3523519992828369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,1,power_law_1.01,0.3634047985076904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,1,power_law_1.01,0.37117440700531007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,1,power_law_1.01,0.3806080102920532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,1,power_law_1.01,0.3943039894104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,1,power_law_1.01,0.39704959392547606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,1,power_law_1.01,0.4149184226989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,1,power_law_1.01,0.41244797706604003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,1,power_law_1.01,0.4219071865081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.5454527854919433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,1,power_law_1.01,0.4328256130218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,1,power_law_1.01,0.4660672187805176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.01,0.4965184211730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.01,0.5820928096771241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.01,0.6511807918548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.01,0.7711679935455322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.01,0.926848030090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.01,1.222815990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.01,1.5351679801940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.01,2.152895927429199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.01,2.7599872589111327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.01,0.11990400552749633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.01,0.11854079961776734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.01,0.11951359510421752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.01,0.1215872049331665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.01,0.11824640035629272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.01,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.01,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.01,0.11862399578094482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.01,0.12183680534362792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.01,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.01,0.1270848035812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.01,0.12858879566192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.01,0.13702399730682374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.01,0.1394495964050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.01,0.14705920219421387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.01,0.1692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.01,0.18398079872131348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.01,0.22983040809631347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.01,0.2562688112258911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.01,0.33498880863189695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.01,0.38762240409851073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.01,0.5073215961456299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.01,0.5994624137878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.01,0.8822976112365722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.01,1.1172991752624513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,power_law_1.01,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,power_law_1.01,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,power_law_1.01,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,power_law_1.01,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,power_law_1.01,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,power_law_1.01,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,power_law_1.01,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.01,0.06519039869308471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.01,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.01,0.12586239576339722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.01,0.1705024003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.01,0.2024768114089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.01,0.2806528091430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.01,0.36160640716552733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,power_law_1.2,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,power_law_1.2,0.0520576000213623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,power_law_1.2,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,power_law_1.2,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,power_law_1.2,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,power_law_1.2,0.07315199971199035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,power_law_1.2,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,power_law_1.2,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,power_law_1.2,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,power_law_1.2,0.10949759483337403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,power_law_1.2,0.12563199996948243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,power_law_1.2,0.156985604763031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,power_law_1.2,0.1884608030319214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,power_law_1.2,0.2620352029800415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.09728639721870422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.12791680097579955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.1734976053237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.21993598937988282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.3204927921295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.3495424032211304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,power_law_1.2,0.3267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.36325759887695314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.3665152072906494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.3727936029434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.38509440422058105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.39792640209198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.3979199886322021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.4159872055053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.42478718757629397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.4442495822906494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4728896141052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5296256065368652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5872831821441651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.7304768085479736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.7700543880462647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.9701567649841308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.1625727653503417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.5248959541320801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,1.8829952239990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,2.607494354248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,3.3778240203857424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,power_law_1.2,0.47098240852355955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,power_law_1.2,0.5997951984405517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,power_law_1.2,0.8167872428894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,balanced,0.030400000512599945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,balanced,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,balanced,0.035989334185918175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,balanced,0.04005333284536997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,balanced,0.08870933453241985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,balanced,0.12866133451461792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,balanced,0.15826666355133057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.01,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.01,0.18673919439315795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.01,0.1832703948020935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.01,0.24109439849853515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.01,0.30601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.01,0.38911359310150145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.01,0.37941761016845704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.01,0.3809727907180786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.01,0.3849280118942261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.01,0.41863040924072265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.01,0.4136256217956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.01,0.42534399032592773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.01,0.4233856201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.01,0.4452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.01,0.45122561454772947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.01,0.4678656101226807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.01,0.49916801452636717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.01,0.5757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.01,0.5549312114715577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.01,0.6585663795471192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.01,0.7291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.01,0.8941184043884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.01,0.9156991958618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.01,1.1475775718688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.01,1.4070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.01,1.9424383163452148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.01,2.2858047485351562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,power_law_1.2,1.1509119987487793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.0743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.0847104012966156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.10094079971313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.11989760398864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.1922047972679138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.2690943956375122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.34734721183776857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.4988736152648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.6388927936553955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.06594560146331788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.06959360241889953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.10035840272903443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.10660480260848999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.14198399782180787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.17217279672622682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.21068799495697021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.24955520629882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.33731200695037844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.6464384078979493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.8098879814147949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.09663360118865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.11674239635467529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.15243519544601442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.19868160486221315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.23967359066009522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.2590143918991089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.3372544050216675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.34639999866485593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.34593920707702636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.35783679485321046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.3721280097961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.37621119022369387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.4102911949157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.41402239799499513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.421830415725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.46494078636169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.48171520233154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.5301824092864991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.6283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.6788095951080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,0.8039744377136231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,0.9606975555419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.2685952186584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,1.5018688201904298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,2.0300031661987306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,2.623980712890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.10502400398254394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.10455679893493652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.029951998591423036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.10706559419631959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.10353920459747315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.11419520378112794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.11425919532775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.11777280569076538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.12273279428482056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.12602239847183228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.1306048035621643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.14847999811172485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.1803264021873474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.20030078887939454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.259935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.29711360931396485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.45199999809265134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,power_law_1.2,0.563040018081665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,power_law_1.2,0.8170816421508789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,power_law_1.2,1.0297216415405273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.1091647982597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.13477760553359985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.1736191987991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.22731521129608154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,balanced,0.06012799839178721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,balanced,0.062128002444903054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,balanced,0.06303999821345012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,balanced,0.0631466656923294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,balanced,0.06448000172773997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06550399959087372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,balanced,0.06457599997520447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,balanced,0.06669333577156067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,balanced,0.07197866837183635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,balanced,0.07603199779987335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,balanced,0.08267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.0906826655069987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.10662399729092915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.129120002190272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.15284799536069235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.17515732844670615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.23332266012827554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.2835306723912557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.39237332344055176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.4940799872080485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.32833280563354494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,power_law_1.01,0.40181760787963866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.06849279999732971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.1285248041152954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.15111680030822755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.17948800325393677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.23412480354309081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.30689918994903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.4618175983428955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,power_law_1.01,0.5466495990753174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,balanced,0.08053333560625713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,balanced,0.08725333213806152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,balanced,0.09685333569844563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,balanced,0.11803733309110005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,balanced,0.1662560005982717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,balanced,0.24724799394607544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,balanced,0.2502666711807251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,balanced,0.25460267066955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,balanced,0.25421865781148273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,balanced,0.25541865825653076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,balanced,0.258735994497935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,balanced,0.264629324277242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,balanced,0.26772799094518024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,balanced,0.2759840091069539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,balanced,0.28364266951878864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,balanced,0.2900266647338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,balanced,0.3047093351682027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,balanced,0.3447360197703044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,balanced,0.37491198380788165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,balanced,0.4508266846338908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,balanced,0.5150613387425741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,balanced,0.6699999968210856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,balanced,0.8224799633026123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,balanced,1.2237066427866619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,balanced,1.5506985982259114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,balanced,2.258783976236979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,balanced,2.9901386896769204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.01,0.12007039785385132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.01,0.18568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.01,0.23228158950805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.01,0.35294721126556394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.01,0.4404287815093994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.01,0.6378687858581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.01,0.7168000221252442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.01,0.7073535919189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.01,0.7052351951599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.01,0.7484799861907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.01,0.7530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.01,0.7836607933044434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.01,0.7898943901062012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.01,0.8002047538757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.01,0.8366527557373047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.01,0.8550527572631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.01,0.9030271530151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.01,0.9735936164855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.01,1.0682944297790526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.01,1.258131217956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.01,1.2755135536193847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.01,1.554528045654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.01,1.6717119216918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.01,2.0420352935791017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.01,2.4710208892822267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.01,3.438585662841797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.01,4.019456100463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.09006080031394958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.11407999992370606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.14200960397720336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.1736575961112976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.2149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.25516159534454347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.372160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,power_law_1.01,0.5245888233184814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,power_law_1.01,0.6907904148101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,balanced,0.05907199780146281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,balanced,0.08130133152008057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,balanced,0.13751999537150064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,balanced,0.13199999928474426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,balanced,0.13241066535313925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,balanced,0.1353653371334076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,balanced,0.13497066497802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,balanced,0.1349066694577535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,balanced,0.13582932949066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,balanced,0.1365066667397817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,balanced,0.13860799868901572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,balanced,0.14130666851997375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,balanced,0.14075733224550882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,balanced,0.1461120049158732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,balanced,0.1551520029703776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,balanced,0.15641599893569946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,balanced,0.17068266868591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,balanced,0.18286399046579996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,balanced,0.2029973268508911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,balanced,0.2493226726849874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,balanced,0.2753066619237264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,balanced,0.36291734377543133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,balanced,0.4404746691385905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,balanced,0.6313333511352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,balanced,0.811743974685669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,balanced,1.1911946932474773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,balanced,1.5582879384358723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,power_law_1.2,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,power_law_1.2,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.09002879858016968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.11618560552597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.13895679712295533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.17155840396881103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.1807935953140259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.18385920524597169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.18869760036468505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.19525760412216187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.20348799228668213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.21305599212646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.22204160690307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.23060479164123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.23773438930511476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.2606911897659302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.3341504096984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.3480063915252686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.44716801643371584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.4857920169830322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6379968166351319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.7753791809082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.0704959869384765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.364793586730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,power_law_1.01,1.9685760498046876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.5385536193847655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.03184640109539032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,power_law_1.2,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.16101759672164917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,power_law_1.2,0.049907198548316954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,power_law_1.2,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.2,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.2,0.10194560289382934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.2,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.2,0.17939200401306152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.2,0.22367360591888427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.18575359582901002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.2,0.31157760620117186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.2,0.40844159126281737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.2,0.6837887763977051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.2,0.9677760124206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,balanced,0.03991466760635376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,balanced,0.03998400022586187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,balanced,0.06271466612815857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,balanced,0.06582933167616527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,balanced,0.06660266717274983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,balanced,0.06646400193373363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,balanced,0.07130666573842366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,balanced,0.07660266757011414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,balanced,0.07870933413505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,balanced,0.08422933022181193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,balanced,0.10191466410954793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,balanced,0.13193600376447043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,balanced,0.15601066748301187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,balanced,0.19970667362213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,balanced,0.23881065845489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,balanced,0.3275946577390035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,balanced,0.41073067982991535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.2720832109451294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,balanced,0.03864533454179764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,balanced,0.08211733400821686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,balanced,0.12660800417264303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,balanced,0.1265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,balanced,0.13184533516565958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,balanced,0.12966932853062949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,balanced,0.1339359978834788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,balanced,0.13563199838002524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,balanced,0.13662399848302206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,balanced,0.14154666662216187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,balanced,0.14152533809343973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,balanced,0.14739200472831726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,balanced,0.15755200386047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,balanced,0.16109333435694376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,balanced,0.1672160029411316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,balanced,0.18931732575098673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,balanced,0.2124320069948832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,balanced,0.26224533716837567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,balanced,0.2948053280512492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,balanced,0.40431467692057294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,balanced,0.47652268409729004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,balanced,0.672165314356486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,balanced,0.849951982498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,balanced,1.2417120138804119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,balanced,1.652330716451009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.339468789100647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,balanced,0.03389866650104523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,balanced,0.03514666606982549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,balanced,0.04529066880544027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.061477333307266235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.08904000123341878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.107013334830602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.1341919998327891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.15524799625078836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.21242133776346842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.26234666506449383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,power_law_1.01,0.02558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,power_law_1.01,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,power_law_1.01,0.024940800666809083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,power_law_1.01,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,power_law_1.01,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,power_law_1.01,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,power_law_1.01,0.02980479896068573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,power_law_1.01,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,power_law_1.01,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,power_law_1.01,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,power_law_1.01,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,power_law_1.01,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,power_law_1.01,0.05516160130500793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,power_law_1.01,0.06600319743156433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,power_law_1.01,0.07694079875946044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,power_law_1.01,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,power_law_1.01,0.1294592022895813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,power_law_1.01,0.19528959989547728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,power_law_1.01,0.25011839866638186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,power_law_1.01,0.3606208086013794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,power_law_1.01,0.4901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,balanced,0.07993066807587941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,balanced,0.10482133428255717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,balanced,0.10629333058993022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,balanced,0.10859733819961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,balanced,0.10700266559918721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,balanced,0.10894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,balanced,0.11083199580510457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,balanced,0.11132799585660298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,balanced,0.11365866661071777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,balanced,0.11434132854143779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,balanced,0.12086932857831319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,balanced,0.12150933345158894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,balanced,0.12519466876983643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,balanced,0.13805866241455078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,balanced,0.14320000012715658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,balanced,0.1648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,balanced,0.1811093290646871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,balanced,0.21639466285705566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,balanced,0.25227200984954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,balanced,0.3452426592508952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,balanced,0.4179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,balanced,0.6010453303654989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,balanced,0.7755040327707926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,power_law_1.2,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,power_law_1.2,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,power_law_1.2,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,power_law_1.2,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,power_law_1.2,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,power_law_1.2,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,power_law_1.2,0.07240319848060608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,power_law_1.2,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,power_law_1.2,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,power_law_1.2,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,power_law_1.2,0.11761280298233032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,power_law_1.2,0.13914239406585693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,power_law_1.2,0.16238720417022706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,power_law_1.2,0.20774400234222412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,power_law_1.2,0.2574336051940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,power_law_1.2,0.378656005859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,power_law_1.2,0.5212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,power_law_1.01,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.01,0.14886399507522582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.01,0.16238080263137816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.01,0.24976639747619628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.01,0.2890304088592529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.07519999742507935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.08997759819030762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.11564160585403442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.16008319854736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.15719679594039918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.17159039974212648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.17226879596710204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.16815999746322632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.1820032000541687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.18655359745025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.20151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.20485119819641112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.21459200382232665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.23257598876953126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.25629439353942873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.31325440406799315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.3044735908508301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.41942400932312013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,power_law_1.2,0.4606912136077881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,power_law_1.2,0.6075712203979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,power_law_1.2,0.7540351867675781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,power_law_1.2,1.0505727767944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,power_law_1.2,1.3985088348388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,power_law_1.2,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,power_law_1.2,0.07242879867553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,power_law_1.2,0.08657280206680298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,power_law_1.2,0.09716479778289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,power_law_1.2,0.09052159786224365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,power_law_1.2,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,power_law_1.2,0.1002303957939148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,power_law_1.2,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,power_law_1.2,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,power_law_1.2,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,power_law_1.2,0.11118719577789307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,power_law_1.2,0.11573760509490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,power_law_1.2,0.11773439645767211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,power_law_1.2,0.12288000583648681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,power_law_1.2,0.13505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,power_law_1.2,0.1412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,power_law_1.2,0.16239999532699584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,power_law_1.2,0.19140479564666749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,power_law_1.2,0.22884480953216552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,power_law_1.2,0.24843521118164064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,power_law_1.2,0.3403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,power_law_1.2,0.3970367908477783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,power_law_1.2,0.5293632030487061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,power_law_1.2,0.7256319999694825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,balanced,0.06071466704209646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,balanced,0.0613919993241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,balanced,0.06570666531721751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,balanced,0.06625066697597504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,balanced,0.06843199829260509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,balanced,0.068122665087382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,balanced,0.08055466910203297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,balanced,0.08637332916259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,balanced,0.09135466814041138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,balanced,0.11152000228563945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,balanced,0.13133866588274637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,balanced,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,balanced,0.20121065775553384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,balanced,0.26633065938949585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,balanced,0.3280533353487651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,balanced,0.4627679983774821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,balanced,0.5883733431498209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.09987840056419373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.12945280075073243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.16763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.23615360260009766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.34202239513397215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,balanced,0.09959466258684795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,balanced,0.1509706676006317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,balanced,0.2528266708056132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,balanced,0.4636319875717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,balanced,0.8795039653778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,balanced,1.0929066340128581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,balanced,1.095626672108968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,balanced,1.096490701039632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,balanced,1.1023306846618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,balanced,1.1062560081481934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,balanced,1.1115307013193767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,balanced,1.1198240121205647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,balanced,1.1253066857655842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,balanced,1.1348053614298503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,balanced,1.1483039855957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,balanced,1.156490643819173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,balanced,1.175872008005778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,balanced,1.254314661026001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,balanced,1.2511306603749592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,balanced,1.4764426549275715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,balanced,1.3886346817016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,balanced,2.070826689402262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,balanced,1.7002453804016113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,balanced,2.5845600763956704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,balanced,2.549525260925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,balanced,4.215216000874837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,balanced,4.648085276285808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,balanced,0.0876693328221639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,balanced,0.08010666569073994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,balanced,0.12821333607037863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,balanced,0.12164266904195149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,balanced,0.12156800429026286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,balanced,0.12195199728012085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,balanced,0.12343466281890869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,balanced,0.1251040001710256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,balanced,0.1253866652647654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,balanced,0.12545599540074667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,balanced,0.12894399960835776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,balanced,0.13570132851600647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,balanced,0.13710400462150574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,balanced,0.1332586705684662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,balanced,0.14940800269444784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,balanced,0.15124799807866415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,balanced,0.1439253290494283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,balanced,0.19877866903940836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,balanced,0.17933867375055948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,balanced,0.23678932587305704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,balanced,0.2516319950421651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,balanced,0.3274986743927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,balanced,0.3752479950586955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,balanced,0.527023990948995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,balanced,0.6900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,balanced,0.9763519763946533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,balanced,1.2798879941304524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,power_law_1.2,0.026495999097824095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,power_law_1.2,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,power_law_1.2,0.026950401067733765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,power_law_1.2,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,power_law_1.2,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,power_law_1.2,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,power_law_1.2,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,power_law_1.2,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,power_law_1.2,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,power_law_1.2,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,power_law_1.2,0.08670719861984252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,power_law_1.2,0.1505087971687317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,power_law_1.2,0.20283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,power_law_1.2,0.31076478958129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,power_law_1.2,0.37832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.2,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.2,0.07002239823341369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.2,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.2,0.07627519965171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.2,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.2,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.2,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.2,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.2,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.2,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.2,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.2,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.2,0.12249599695205689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.2,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.2,0.15762560367584227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.2,0.1687551975250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.2,0.22481920719146728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.2,0.26145920753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.2,0.3477312088012695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.2,0.4305600166320801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.2,0.5641600131988526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,power_law_1.2,0.5730944156646729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.2,0.7105216026306153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,power_law_1.2,0.8804032325744628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.08614400029182434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.09772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.11857919692993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.12116479873657227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.14547200202941896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.15247360467910767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.229913592338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.27080960273742677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,power_law_1.2,0.3477567911148071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,power_law_1.2,0.39927680492401124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,power_law_1.2,0.5821951866149903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,power_law_1.2,0.8677184104919433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,balanced,0.07303999861081441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,balanced,0.11391466856002808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,balanced,0.11499733726183574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,balanced,0.11473600069681804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,balanced,0.11608533064524333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,balanced,0.11564800143241882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,balanced,0.11690133810043335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,balanced,0.11894399921099345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,balanced,0.11841600139935811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,balanced,0.12427199880282085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,balanced,0.12505599856376648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,balanced,0.1269439955552419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,balanced,0.1418400009473165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,balanced,0.14195199807484946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,balanced,0.16520532965660095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,balanced,0.16853866974512735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,balanced,0.22585066159566244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,balanced,0.23583465814590454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,balanced,0.3574560085932414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,balanced,0.35441064834594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,balanced,0.47886931896209717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,balanced,0.6138773361841837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.07089920043945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.10353280305862426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.12101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.1380735993385315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.1837504029273987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.2236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.31884160041809084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.3732480049133301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,power_law_1.2,0.5642752170562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,power_law_1.2,0.7958144187927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,power_law_1.2,0.11215360164642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,power_law_1.2,0.10866559743881225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,power_law_1.2,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,power_law_1.2,0.1097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,power_law_1.2,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,power_law_1.2,0.09850879907608032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,power_law_1.2,0.10684800148010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,power_law_1.2,0.10525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,power_law_1.2,0.10915839672088623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,power_law_1.2,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,power_law_1.2,0.11192320585250855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,power_law_1.2,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,power_law_1.2,0.11441279649734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,power_law_1.2,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,power_law_1.2,0.12536959648132323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,power_law_1.2,0.12266240119934083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,power_law_1.2,0.13916800022125245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,power_law_1.2,0.1609663963317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.2,0.1698815941810608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.2,0.2017535924911499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.2,0.2477504014968872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.2,0.29845759868621824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.2,0.3710783958435059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.2,0.548364782333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.2,0.7517824172973633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.2,1.029759979248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.2,1.318272018432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,balanced,0.03381866713364919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,balanced,0.03346666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,balanced,0.03383466601371765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,balanced,0.05162133276462555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,balanced,0.0736053337653478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,balanced,0.09684266646703084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,balanced,0.10918933153152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,balanced,0.144378662109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,balanced,0.1677280068397522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.06481279730796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.12038400173187255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.14954880475997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.17320319414138793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.22421760559082032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.27253758907318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.3769344091415405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,power_law_1.01,0.02107519954442978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,power_law_1.01,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,power_law_1.01,0.023231999576091768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,power_law_1.01,0.02454400062561035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,power_law_1.01,0.02481919974088669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,power_law_1.01,0.02598400115966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,power_law_1.01,0.028262400627136232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,power_law_1.01,0.02898559868335724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,power_law_1.01,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,power_law_1.01,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,power_law_1.01,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.08769919872283935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.11023999452590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.14331519603729248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,balanced,0.04347200194994608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,balanced,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,balanced,0.044213334719340004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,balanced,0.04548266530036926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,balanced,0.048672000567118325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,balanced,0.06640000144640605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,balanced,0.06701333324114482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,balanced,0.08001066744327545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,balanced,0.08928533395131429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.11750933527946472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.1388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.17904533942540488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.21485867102940878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.30001066128412884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.3811093171437581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.5485333204269409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.7147680123647054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.47806081771850584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.6790272235870362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.8838848114013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.04010879993438721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.05692800283432007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.07063680291175842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.0964735984802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.11573760509490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.12697600126266478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.15412479639053345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.20280320644378663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.2622463941574097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.3782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.4647871971130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.6660096168518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.01,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.01,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.01,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.01,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.01,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.01,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.01,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.01,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.01,0.08504319787025452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.01,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.01,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.01,0.12040959596633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.01,0.12586239576339722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.01,0.16004480123519899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.01,0.17450239658355712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.01,0.2244096040725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.01,0.24249598979949952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.01,0.3276544094085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.01,0.4024511814117432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.01,0.5570432186126709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.01,0.6574783802032471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.14481279850006104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.24861440658569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.33980159759521483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.3342335939407349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.33803520202636717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.3145792007446289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.26985599994659426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.29073920249938967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.3128895998001099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.29298560619354247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.2575167894363403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.3134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,power_law_1.01,0.06420480012893677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,power_law_1.01,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,power_law_1.01,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,power_law_1.01,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,power_law_1.01,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.2987456083297729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.3101311922073364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.3191551923751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.30364160537719725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.35015039443969725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,power_law_1.01,0.09902719855308532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,power_law_1.01,0.10676480531692505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.36613759994506834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.40648322105407714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.48472962379455564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,power_law_1.01,0.12579200267791749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.4924736022949219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,power_law_1.01,0.1541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,power_law_1.2,0.6204544067382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,power_law_1.2,0.6962175846099854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,power_law_1.01,0.1828927993774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,power_law_1.2,0.8836928367614746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,power_law_1.2,1.0014016151428222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,power_law_1.2,1.244326400756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,power_law_1.2,1.727289581298828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,power_law_1.01,0.225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,power_law_1.01,0.26380798816680906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,power_law_1.01,0.38461439609527587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,power_law_1.01,0.4237504005432129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,power_law_1.01,0.76212477684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,balanced,0.029557332396507263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,balanced,0.03173866619666418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,balanced,0.03436266630887985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,power_law_1.01,0.8982975959777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,balanced,0.04298133154710134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,balanced,0.05892266829808553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,balanced,0.06462933123111725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,balanced,0.09079466263453166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,balanced,0.11748799681663513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,balanced,0.14009599884351095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,balanced,0.1900320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,balanced,0.23666133483250937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,power_law_1.01,1.3740927696228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,power_law_1.01,1.916249656677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,balanced,0.07584533095359802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,balanced,0.12081600228945415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,balanced,0.20292800664901733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,balanced,0.20706133047739664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,balanced,0.20754667123158774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,balanced,0.20868800083796182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,balanced,0.21071465810139975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,balanced,0.2121386726697286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,balanced,0.21239999930063883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,balanced,0.21473066012064615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,balanced,0.22070932388305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,balanced,0.22378132740656534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,balanced,0.22422399123509726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,balanced,0.2330346703529358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,balanced,0.25154133637746173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,balanced,0.25221866369247437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,balanced,0.25827733675638836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,balanced,0.3160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,balanced,0.3307360013326009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,balanced,0.406821330388387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,balanced,0.46883734067281085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,balanced,0.6341813405354818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,balanced,0.7940959930419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,balanced,1.1492640177408855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,balanced,1.4966932932535808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,balanced,2.201061407725016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,balanced,2.92415459950765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05516160130500793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.08311039805412293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.10581120252609252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.10883200168609619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.11263999938964844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.1155135989189148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.1236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.13406720161437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.1394368052482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.1490623950958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.1697216033935547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.17968000173568727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.2041088104248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.2463167905807495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.32654080390930174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.40842242240905763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,power_law_1.2,0.5237311840057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,power_law_1.2,0.6880896091461182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,power_law_1.2,1.037343978881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,power_law_1.2,1.2178624153137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,balanced,0.06049066781997681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,balanced,0.06251733501752217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,balanced,0.06262399752934773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,balanced,0.06328533093134563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,balanced,0.06373866895834605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,balanced,0.0726560006539027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,balanced,0.07333333293596904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,balanced,0.07477333148320515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,balanced,0.09144000212351482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,balanced,0.13153066237767538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,balanced,0.14323733250300089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,balanced,0.19810134172439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,balanced,0.2404693365097046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,balanced,0.3460906744003296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,balanced,0.44201068083445233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,balanced,0.6547093391418457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,balanced,0.8534933725992838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,power_law_1.01,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,power_law_1.01,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,power_law_1.01,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,power_law_1.01,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,power_law_1.01,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,power_law_1.01,0.08334720134735107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,power_law_1.01,0.09972479939460754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,power_law_1.01,0.1085312008857727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,power_law_1.01,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,power_law_1.01,0.14145920276641846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,power_law_1.01,0.16792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,power_law_1.01,0.21413118839263917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,power_law_1.01,0.28408958911895754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,power_law_1.01,0.3480191946029663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,power_law_1.01,0.4800896167755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,power_law_1.01,0.6377280235290528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,power_law_1.01,0.8828351974487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,power_law_1.01,1.1940799713134767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.11088000535964966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.11189759969711303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.1031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.10567679405212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.10864640474319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.10782719850540161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.1072383999824524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.11112960577011108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.11568000316619872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.12099200487136841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.1214143991470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.1308608055114746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.15411200523376464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.17970559597015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.20033280849456786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.24335360527038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.2876159906387329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.3946559906005859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.45834879875183104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.6436607837677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.8643263816833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.11318399906158447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.17964160442352295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.23573119640350343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.07672320008277893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.0985472023487091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.1230847954750061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.21738240718841553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.2371583938598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.24634881019592286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.24865920543670655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.2571135997772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.26574718952178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.2717375993728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.2803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.2846528053283691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.29923200607299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.30821120738983154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.3275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.3775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.41237759590148926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.5131711959838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.5878015995025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,0.7664383888244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,0.933011245727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,1.2690048217773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,1.7545087814331055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,2.4129472732543946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,2.9848384857177734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,power_law_1.01,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,power_law_1.01,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,power_law_1.01,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,power_law_1.01,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,power_law_1.01,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,power_law_1.01,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,power_law_1.01,0.09624320268630981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,power_law_1.01,0.11288319826126099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,power_law_1.01,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,power_law_1.01,0.20087039470672607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,power_law_1.01,0.2742975950241089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,power_law_1.01,0.3207040071487427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,power_law_1.01,0.46161279678344724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,power_law_1.01,0.6128191947937012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.049907198548316954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.05568640232086182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.05809280276298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.06958079934120179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.0953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.11147520542144776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.13459839820861816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.1646783947944641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.1997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.26691200733184817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.33870720863342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.4950592041015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.6662655830383301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,power_law_1.2,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,power_law_1.2,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,power_law_1.2,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,power_law_1.2,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,power_law_1.2,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,power_law_1.2,0.08071039915084839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,power_law_1.2,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,power_law_1.2,0.1095039963722229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,power_law_1.2,0.12929919958114625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,power_law_1.2,0.16512000560760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,power_law_1.2,0.21584639549255372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,power_law_1.2,0.27915520668029786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,power_law_1.2,0.35859839916229247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,power_law_1.2,0.5294271945953369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,power_law_1.2,0.6894847869873046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.15637760162353515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.25107200145721437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.4349952220916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.7582975864410401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.0467583656311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.267903995513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.5987199783325194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.6598272323608398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.7194047927856446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.7722944259643554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.8183168411254882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,power_law_1.01,1.8960575103759765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,power_law_1.01,1.9342336654663086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.0312192916870115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.057548713684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.1146112442016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.2160959243774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.376198387145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,power_law_1.01,2.627827262878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,power_law_1.01,2.6478656768798827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,power_law_1.01,2.9757055282592773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,power_law_1.01,3.488159942626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,power_law_1.01,4.4100288391113285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,power_law_1.01,4.251148986816406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,power_law_1.01,5.297951889038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,power_law_1.01,6.856179046630859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.02683520019054413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.028531199693679808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.028441599011421202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.02937600016593933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.04993920028209686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.10888960361480712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.11866240501403809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.16020480394363404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.18410880565643312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,power_law_1.01,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,power_law_1.01,0.08202239871025085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,power_law_1.01,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,power_law_1.01,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,power_law_1.01,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,power_law_1.01,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,power_law_1.01,0.09313920140266418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,power_law_1.01,0.09374719858169556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,power_law_1.01,0.09462400078773499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,power_law_1.01,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,power_law_1.01,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,power_law_1.01,0.11709439754486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,power_law_1.01,0.1323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.01,0.14260480403900147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.01,0.164902400970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.01,0.1956544041633606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.01,0.2479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.01,0.3295039892196655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.01,0.44572157859802247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.01,0.5979008197784423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.01,0.9186367988586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.01,1.1461376190185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,balanced,0.06446933249632518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,balanced,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,balanced,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,balanced,0.2032053271929423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,balanced,0.33713066577911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,balanced,0.4498293399810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,balanced,0.45163734753926593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,balanced,0.4557493527730306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,balanced,0.4561973412831624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,balanced,0.4585706790288289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,balanced,0.46164266268412274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,balanced,0.4649866819381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,balanced,0.46714667479197186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,balanced,0.47113601366678876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,balanced,0.48187732696533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,balanced,0.48317865530649823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,balanced,0.492634654045105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,balanced,0.5989973147710165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,balanced,0.5344800154368082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,balanced,0.6728160381317139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,balanced,0.6260213454564413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,balanced,0.9834293524424235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,balanced,0.7741173108418783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,balanced,1.2014293670654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,balanced,1.1677227020263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,balanced,1.735498587290446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,balanced,1.878453254699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.2772608041763306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,power_law_1.2,0.3559103965759277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.09276800155639649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.10040960311889649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.1124735951423645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.13422720432281493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.1220352053642273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.1395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.1353279948234558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.14632320404052734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.1492735981941223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.1651136040687561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.1918272018432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.20268800258636474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.24355840682983398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.2761791944503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.34558079242706297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.44855680465698244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.574067211151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,power_law_1.2,0.6989183902740479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,power_law_1.2,0.9306943893432618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,power_law_1.2,1.2629568099975585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,power_law_1.01,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,power_law_1.01,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,power_law_1.01,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,power_law_1.01,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,power_law_1.01,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,power_law_1.01,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,power_law_1.01,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,power_law_1.01,0.12197120189666748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,power_law_1.01,0.15482879877090455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,power_law_1.01,0.18746880292892457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,power_law_1.01,0.24729599952697753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,power_law_1.01,0.35595519542694093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,power_law_1.01,0.4593344211578369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,power_law_1.01,0.6021632194519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,power_law_1.2,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,power_law_1.2,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,power_law_1.2,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,power_law_1.2,0.08783360123634339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,power_law_1.2,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,power_law_1.2,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,power_law_1.2,0.09582080245018006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,power_law_1.2,0.10142719745635986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,power_law_1.2,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,power_law_1.2,0.10831359624862671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,power_law_1.2,0.1139456033706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,power_law_1.2,0.11915520429611207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,power_law_1.2,0.12485120296478272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,power_law_1.2,0.14947839975357055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,power_law_1.2,0.15479040145874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,power_law_1.2,0.19923839569091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,power_law_1.2,0.21585919857025146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,power_law_1.2,0.27552640438079834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,power_law_1.2,0.32111999988555906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,power_law_1.2,0.47146239280700686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,power_law_1.2,0.6136320114135743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,power_law_1.2,0.8728256225585938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,power_law_1.2,1.1191360473632812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.2,0.18755840063095092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.2,0.2478463888168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.2,0.3178112030029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.2,0.4421567916870117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.2,0.5357120037078857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.2,0.6854591846466065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.2,0.7064256191253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.2,0.7526144027709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.2,0.7491583824157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.2,0.7362112045288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.2,0.7750976085662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.2,0.7831552028656006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,power_law_1.2,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,power_law_1.2,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.2,0.805958366394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.2,0.837343978881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,power_law_1.2,0.058457601070404056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,power_law_1.2,0.062080001831054686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.2,0.8583616256713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,power_law_1.2,0.05976960062980652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.2,0.8505855560302734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.2,0.9280832290649415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,power_law_1.2,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.2,1.0547519683837892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.2,1.0385855674743651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,power_law_1.2,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.2,1.188748836517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,power_law_1.2,0.11191040277481079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.2,1.268992042541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.2,1.5270719528198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.2,1.6560319900512694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.2,2.000294494628906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,power_law_1.2,0.11854079961776734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.2,2.4954559326171877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,power_law_1.2,0.1477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.2,3.392652893066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,power_law_1.2,0.17262719869613646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,power_law_1.2,0.23374080657958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.2,3.724300765991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,power_law_1.2,0.2834176063537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.09194880127906799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.0917952001094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.09211519956588746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.11276160478591919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.11971839666366577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.130457603931427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.1491328001022339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.17470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.20451838970184327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.25102078914642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.2944447994232178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.4100031852722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.47481598854064944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.6519296169281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.8934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,power_law_1.2,0.40315518379211424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,power_law_1.2,0.5135935783386231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,power_law_1.2,0.7353216171264648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.1524415969848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.24078080654144288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.4371647834777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.6274176120758057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.8506303787231445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.1704895973205567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.217471981048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.2616640090942384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.2740159988403321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.3308223724365233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.352774429321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.404543972015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.429094409942627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.4878463745117188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.4918784141540526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.5734911918640138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.6794879913330079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,power_law_1.01,1.678860855102539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,power_law_1.2,0.9508159637451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,power_law_1.01,1.8641471862792969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,power_law_1.01,1.9398719787597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,power_law_1.01,2.2290687561035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,power_law_1.01,2.474847984313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,power_law_1.01,2.954015922546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,power_law_1.01,3.3746688842773436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,power_law_1.01,4.347423934936524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,power_law_1.01,5.420185470581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,power_law_1.2,0.06512640118598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,power_law_1.2,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,power_law_1.2,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,power_law_1.2,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,power_law_1.2,0.08830080032348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,power_law_1.2,0.08932480216026306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,power_law_1.2,0.0908415973186493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,power_law_1.2,0.10153599977493286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,power_law_1.2,0.11090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,power_law_1.2,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,power_law_1.2,0.12186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,power_law_1.2,0.13660800457000732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,power_law_1.2,0.17111680507659913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.2,0.19997440576553344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.2,0.2603775978088379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.2,0.2880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.2,0.4077568054199219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.2,0.5590464115142822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.2,0.7539904117584229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,power_law_1.2,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,power_law_1.2,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,power_law_1.2,0.12936960458755492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,power_law_1.2,0.15940480232238768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,power_law_1.2,0.2575743913650513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,power_law_1.2,0.27603840827941895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,power_law_1.2,0.2868736028671265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,power_law_1.2,0.29706881046295164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,power_law_1.2,0.30390400886535646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,power_law_1.2,0.3155839920043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,power_law_1.2,0.3291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,power_law_1.2,0.3236736059188843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,power_law_1.2,0.3401472091674805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,power_law_1.2,0.3634752035140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,power_law_1.2,0.3763711929321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,power_law_1.2,0.39423360824584963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,power_law_1.2,0.43709440231323243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,power_law_1.2,0.4838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,power_law_1.2,0.5886847972869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,power_law_1.2,0.7064127922058105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,power_law_1.2,0.9150143623352051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,power_law_1.2,1.039072036743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,power_law_1.2,1.3569536209106445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.2,1.0088704109191895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,power_law_1.2,1.7539840698242188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,power_law_1.2,2.424857521057129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,power_law_1.2,3.0965503692626952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.2,1.70579833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.2,1.7991615295410157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,power_law_1.2,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,power_law_1.2,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,power_law_1.2,0.1291584014892578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,power_law_1.2,0.12263679504394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,power_law_1.2,0.13111679553985595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,power_law_1.2,0.13684480190277098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,power_law_1.2,0.14027520418167114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,power_law_1.2,0.1334720015525818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,power_law_1.2,0.1415488004684448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,power_law_1.2,0.1420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,power_law_1.2,0.15525120496749878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,power_law_1.2,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,power_law_1.2,0.1726591944694519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,power_law_1.2,0.17751679420471192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,power_law_1.2,0.19456000328063966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,power_law_1.2,0.2153856039047241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,power_law_1.2,0.26511359214782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,power_law_1.2,0.2896960020065308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,power_law_1.2,0.36424961090087893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,power_law_1.2,0.4244351863861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,power_law_1.2,0.5763775825500488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,power_law_1.2,0.756063985824585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,power_law_1.2,0.9999103546142578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,power_law_1.2,1.317535972595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,balanced,0.03607466568549474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,balanced,0.036730666955312095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,balanced,0.05239999790986379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,balanced,0.07206933200359344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,balanced,0.07782933115959167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,balanced,0.09673066933949788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,balanced,0.11044800281524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,balanced,0.14103466272354126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,balanced,0.1712906757990519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,balanced,0.230512003103892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,balanced,0.28969067335128784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,power_law_1.2,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,power_law_1.2,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,power_law_1.2,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,power_law_1.2,0.06481279730796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,power_law_1.2,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,power_law_1.2,0.06533759832382202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,power_law_1.2,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,power_law_1.2,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,power_law_1.2,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,power_law_1.2,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,power_law_1.2,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,power_law_1.2,0.1261952042579651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.2,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.2,0.1920896053314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.2,0.21448960304260253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.2,0.3287487983703613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.2,0.37664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.2,0.52259840965271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,power_law_1.2,0.0743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,power_law_1.2,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,power_law_1.2,0.09669119715690613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,power_law_1.2,0.11023999452590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,power_law_1.2,0.13014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,power_law_1.2,0.15539200305938722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,power_law_1.2,0.16154240369796752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,power_law_1.2,0.1640895962715149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,power_law_1.2,0.16773760318756104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,power_law_1.2,0.17329280376434325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,power_law_1.2,0.17991679906845093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,power_law_1.2,0.18379520177841185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,power_law_1.2,0.19057919979095458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,power_law_1.2,0.1993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,power_law_1.2,0.206931209564209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,power_law_1.2,0.2217087984085083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,power_law_1.2,0.2572736024856567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.2,0.28431360721588134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.2,0.3517823934555054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.2,0.41324801445007325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.2,0.5412735939025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.2,0.6669695854187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.2,0.927187156677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.2,1.1820223808288575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.2,1.6918336868286132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.2,2.2193792343139647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.2,0.7422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.2,1.0234687805175782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.04190079867839813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.061977601051330565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.1287935972213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.16145919561386107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.063673597574234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.066975998878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.107315194606781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.13393919467926024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.15803519487380982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.20721919536590577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.23915519714355468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.32026240825653074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.39957120418548586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.5564095973968506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.7331647872924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.21848959922790528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.2,1.508236789703369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.26886401176452634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.36348159313201905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,power_law_1.01,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,power_law_1.01,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,power_law_1.01,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,power_law_1.01,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.17760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.4397247791290283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,balanced,0.05453333258628845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,balanced,0.07401599983374278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,balanced,0.10300800204277039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,balanced,0.14128533005714417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,balanced,0.1803413430849711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,balanced,0.18388799826304117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,balanced,0.1835093299547831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,balanced,0.18386665980021158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,balanced,0.188591996828715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,balanced,0.19002666076024374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,balanced,0.19188799460728964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,balanced,0.18954666455586752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,balanced,0.19568532705307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,balanced,0.1974560022354126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,balanced,0.20962133010228476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,balanced,0.22502400477727255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.23628799120585123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.26665065685908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.29179733991622925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.34784531593322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.40453867117563885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.5237760146458944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.6852533022562662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,balanced,0.9130132993062338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,balanced,1.2073386510213215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,power_law_1.2,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,power_law_1.2,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,power_law_1.2,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,power_law_1.2,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,power_law_1.2,0.08273919820785522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,power_law_1.2,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,power_law_1.2,0.09989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,power_law_1.2,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,power_law_1.2,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,power_law_1.2,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.178547203540802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.19895039796829223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2626176118850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.3285056114196777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.4479936122894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.5668543815612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.7969727993011475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.01,0.12319999933242798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.01,0.18914560079574586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.01,0.2250368118286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.01,0.3579200029373169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.01,0.45326080322265627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.01,0.6638783931732177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.01,0.6699647903442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.01,0.714246416091919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.01,0.7441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.01,0.7525824069976806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.01,0.7648384094238281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.01,0.7966784000396728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.01,0.7762368202209473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.01,0.8096768379211425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.01,0.8489536285400391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.01,0.860416030883789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.01,0.9286848068237304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.01,1.0385024070739746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.01,1.0226431846618653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.01,1.1733887672424317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.01,1.3202367782592774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.01,1.5796287536621094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.01,1.5908991813659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.01,1.9521024703979493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.01,2.4025087356567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.01,3.4360511779785154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.01,3.773004913330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.0302656173706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13772159814834595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.14132479429244996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.14035840034484864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.14499839544296264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.14748159646987916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.14828799962997435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.1547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.11615999937057495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.15859839916229249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.16162559986114503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.13600640296936034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.1773568034172058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.14724479913711547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.18193279504776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.14794880151748657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.15327359437942506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.16165119409561157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.20206079483032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.17084800004959105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.16748160123825073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.23331840038299562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.17715840339660643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.2566400051116943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.19467519521713256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.21779201030731202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.2458048105239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.30662400722503663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.35810561180114747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.4365375995635986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.5843200206756591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.7820799827575684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,power_law_1.2,0.957094383239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.3243391990661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.311302375793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,power_law_1.2,1.6490304946899415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.39051520824432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.5182079792022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,balanced,0.044256001710891724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,balanced,0.0610346645116806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.07607466479142506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.08377066254615784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.09922132889429729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.11154666543006897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.13858133554458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.1630453368028005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.21982399622599283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.26951466004053753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.6288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.8658880233764649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,balanced,0.0532533327738444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,balanced,0.05316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,balanced,0.06500266492366791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,balanced,0.06646400193373363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,balanced,0.07243200143178304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,balanced,0.08518933256467183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,balanced,0.09456533193588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,balanced,0.10940800110499065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,balanced,0.12583466370900473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,balanced,0.15479466319084167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,balanced,0.18650666872660318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,balanced,0.2469759980837504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,balanced,0.3027893304824829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,balanced,0.42979200681050617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,balanced,0.5441333452860514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.0761088371276855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.5262911796569825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,1.9686975479125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.20802559852600097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.21704959869384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.2591808080673218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.27198081016540526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.2776319980621338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.2842751979827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.29982719421386717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.30779519081115725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.3109247922897339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.33152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.33875839710235595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.3423295974731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.2,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.3817663908004761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.4158656120300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.2,0.10611840486526489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.4913856029510498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.2,0.13948160409927368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.5346687793731689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.2,0.1916159987449646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.639686393737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,power_law_1.01,0.7812543869018554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.2,0.23471360206604003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.2,0.3575615882873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.0266112327575683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.2,0.391379189491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.2845952033996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.2,0.3854527950286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,power_law_1.01,1.6855808258056642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.2,0.3968447923660278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.2,0.40901761054992675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.2,0.4225855827331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,power_law_1.01,2.0769279479980467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.2,0.43169918060302737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.2,0.45204482078552244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.2,0.45729918479919435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.2,0.4714816093444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.2,0.4858687877655029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.2,0.49989118576049807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.2,0.5626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.2,0.6043647766113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.2,0.7236800193786621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.2,0.7976319789886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.2,1.0039551734924317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.2,1.0580544471740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.2,1.3963775634765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.2,1.8133888244628906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.2,2.559449577331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.2,3.0915584564208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,power_law_1.2,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,power_law_1.2,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,power_law_1.2,0.10983040332794189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,power_law_1.2,0.11461119651794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,power_law_1.2,0.11525119543075561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,power_law_1.2,0.11671680212020874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,power_law_1.2,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,power_law_1.2,0.12033280134201049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,power_law_1.2,0.12258559465408325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,power_law_1.2,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,power_law_1.2,0.12899839878082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,power_law_1.2,0.13218560218811035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,power_law_1.2,0.1382848024368286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,power_law_1.2,0.14577280282974242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,power_law_1.2,0.1691264033317566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.2,0.18491519689559938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.2,0.23737599849700927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.2,0.2971519947052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.2,0.3879296064376831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.2,0.4938047885894775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.2,0.699564790725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.2,0.9386176109313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.2,1.3388416290283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.2,0.09675520062446594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.2,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.2,0.10369280576705933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.2,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.2,0.10350079536437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.2,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.2,0.10575360059738159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.2,0.10405119657516479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.2,0.10373120307922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.2,0.10650240182876587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.2,0.11159039735794067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.2,0.12376960515975952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.2,0.1268224000930786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.2,0.13217920064926147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.2,0.16565760374069213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.2,0.17538559436798096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.2,0.21969919204711913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.2,0.23191680908203124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.2,0.30576000213623045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.2,0.3214143991470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.2,0.4355008125305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.2,0.6388224124908447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.2,0.8174783706665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.2,0.8914303779602051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.2,1.9364095687866212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,power_law_1.2,0.027379199862480164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,power_law_1.2,0.02651520073413849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,power_law_1.2,0.027046400308609008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,power_law_1.2,0.02821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,power_law_1.2,0.029254400730133058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,power_law_1.2,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,power_law_1.2,0.06861439943313599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,power_law_1.2,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,power_law_1.2,0.12183680534362792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,power_law_1.2,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,power_law_1.2,0.20606720447540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,power_law_1.2,0.26458239555358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,power_law_1.2,0.4317056179046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,balanced,0.025445332129796345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,balanced,0.02553066611289978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,balanced,0.029637334247430164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,balanced,0.027301333844661713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,balanced,0.027285332481066387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,balanced,0.030975999931494396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,balanced,0.03130666663249334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,balanced,0.035205334424972534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,balanced,0.03528533379236857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,balanced,0.07906666894753774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,power_law_1.2,0.5519487857818604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,balanced,0.053599998354911804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,balanced,0.06830400228500366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,balanced,0.07238933444023132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.10100799798965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.12685867150624594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.14777599771817526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.19563732544581094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.23993066946665445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.3309866587320964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.41250133514404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,balanced,0.036703998843828835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,balanced,0.035989334185918175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,balanced,0.03938133269548416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.0496373325586319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.07935999830563863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.11978666981061299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.1814346710840861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.133024001121521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.132697594165802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.1371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.14165120124816893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.1431872010231018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.14495999813079835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.15229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.15908479690551758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.15934079885482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.16451200246810913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.18267519474029542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.1791103959083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.20690560340881348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.22231678962707518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.2544895887374878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.3007551908493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,power_law_1.01,0.36212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,power_law_1.01,0.5063936233520507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,power_law_1.01,0.5639616012573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,power_law_1.01,0.767903995513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,power_law_1.01,1.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,power_law_1.01,1.6293119430541991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,power_law_1.01,2.2676095962524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,balanced,0.06632533172766368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,balanced,0.06923200190067291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,balanced,0.06929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,balanced,0.06851733227570851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,balanced,0.07055466870466869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,balanced,0.06984533369541168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,balanced,0.07015466690063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,balanced,0.07167466481526692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,balanced,0.07739733159542084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,balanced,0.08386666576067607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,balanced,0.08866133292516072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,balanced,0.09776533643404643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,balanced,0.11145599683125813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,balanced,0.14189866185188293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,balanced,0.15995200475056967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,balanced,0.21976532538731894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,balanced,0.27565866708755493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,balanced,0.40462398529052734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,balanced,0.5214026769002279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,balanced,0.04823466638724009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,balanced,0.07580266892910004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,balanced,0.07683200140794118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,balanced,0.07856533428033192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,balanced,0.07713599999745686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,balanced,0.07899733384450276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,balanced,0.0784853349129359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,balanced,0.08088533580303192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,balanced,0.09502933422724406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,balanced,0.10027199983596802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,balanced,0.11122666796048482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,balanced,0.11365866661071777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,balanced,0.13165332873662314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,balanced,0.14375999569892883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,balanced,0.17750932772954306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,balanced,0.20277865727742514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,balanced,0.2708746592203776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,balanced,0.334175984064738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,balanced,0.47626666227976483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,balanced,0.6321599880854288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,balanced,0.06497066716353099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,balanced,0.07042666773001353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,balanced,0.06845866640408833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,balanced,0.06911466519037883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,balanced,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,balanced,0.07095466554164886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,balanced,0.07274133463700612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,balanced,0.07628799974918365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,balanced,0.08091733356316884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,balanced,0.08246933420499165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,balanced,0.09821333487828572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,balanced,0.10603200395901997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,balanced,0.13517866532007852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,balanced,0.14801067113876343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,balanced,0.21013865868250528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,balanced,0.25115732351938885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,balanced,0.3552960157394409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,balanced,0.43806934356689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,balanced,0.632202665011088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,balanced,0.8144480387369791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,power_law_1.2,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,power_law_1.2,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,power_law_1.2,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,power_law_1.2,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,power_law_1.2,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,power_law_1.2,0.08408960103988647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,power_law_1.2,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,power_law_1.2,0.08968960046768189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,power_law_1.2,0.09322879910469055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,power_law_1.2,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,power_law_1.2,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,power_law_1.2,0.10967680215835571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,power_law_1.2,0.1230847954750061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,power_law_1.2,0.1375488042831421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,power_law_1.2,0.15936000347137452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,power_law_1.2,0.19866880178451538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,power_law_1.2,0.23189120292663573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,power_law_1.2,0.30494720935821534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,power_law_1.2,0.363916802406311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,power_law_1.2,0.5219840049743653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,power_law_1.2,0.6844672203063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,power_law_1.2,1.0191871643066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,power_law_1.2,1.2472319602966309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,power_law_1.2,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,power_law_1.2,0.06336640119552613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,power_law_1.2,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,power_law_1.2,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,power_law_1.2,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,power_law_1.2,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,power_law_1.2,0.09571200013160705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,power_law_1.2,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,power_law_1.2,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,power_law_1.2,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,power_law_1.2,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,power_law_1.2,0.12677760124206544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,power_law_1.2,0.14440319538116456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.2,0.1557952046394348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.2,0.1870911955833435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.2,0.22042880058288575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.2,0.2845632076263428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.2,0.35751678943634035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.2,0.49752960205078123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.2,0.6504576206207275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.2,0.9167872428894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.2,1.2209792137145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,balanced,0.03291733314593633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,balanced,0.03668266783157984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,balanced,0.05406933526198069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,balanced,0.058602665861447654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,balanced,0.06397866706053416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,balanced,0.06774400174617767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,balanced,0.07218666871388753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,balanced,0.07946666578451793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,balanced,0.10020800431569417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,balanced,0.11057066917419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,balanced,0.1465120017528534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,balanced,0.1689866582552592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,balanced,0.2348960041999817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,balanced,0.29043734073638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,power_law_1.2,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,power_law_1.2,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,power_law_1.2,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,power_law_1.2,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,power_law_1.2,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,power_law_1.2,0.06961280107498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,power_law_1.2,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,power_law_1.2,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,power_law_1.2,0.09004799723625183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,power_law_1.2,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,power_law_1.2,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,power_law_1.2,0.1448192000389099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,power_law_1.2,0.20127360820770263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,power_law_1.2,0.2437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,power_law_1.2,0.35288960933685304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,balanced,0.04146666576464971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,balanced,0.04933333396911621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,balanced,0.06819200019041698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,balanced,0.07300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,power_law_1.2,0.4579775810241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,balanced,0.08693333466847737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,balanced,0.09923199812571208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,balanced,0.1342026690642039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,balanced,0.15954666336377463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,balanced,0.21273066600163779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,balanced,0.26521066824595135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.2,0.1371135950088501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.2,0.1902783989906311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,power_law_1.2,0.6599232196807862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.2,0.32539520263671873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.2,0.48215041160583494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.2,0.6574719905853271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.2,1.2038399696350097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.2,1.28603515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.2,1.3485247611999511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.2,1.3846912384033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.2,1.4350015640258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.2,1.4815168380737305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.2,1.5272704124450684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.2,1.5738880157470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.2,1.636332893371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.2,1.676595115661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.2,1.7141311645507813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.2,1.8188287734985351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.2,2.051807975769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.2,1.9769279479980468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.2,2.2558719635009767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.2,2.4328639984130858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.2,2.8694080352783202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.2,3.040934371948242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.2,3.687846374511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.2,4.531699371337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.2,6.156076812744141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.2,7.025785827636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,power_law_1.2,0.826848030090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,balanced,0.03589333345492681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,balanced,0.03425066669782003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,balanced,0.031199999153614044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,balanced,0.0373279998699824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,balanced,0.036687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,balanced,0.041482667128245033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,balanced,0.04496533175309499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,balanced,0.07162666817506154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,balanced,0.08571199576059978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,balanced,0.10218666990598042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,balanced,0.12160533666610718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,balanced,0.15103466312090555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,power_law_1.2,1.1535103797912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,power_law_1.2,1.6443008422851562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,balanced,0.056421334544817604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,balanced,0.04236799975236257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,balanced,0.04428799947102865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,balanced,0.04738133152325948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,balanced,0.04689066608746847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,balanced,0.04839999973773956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,balanced,0.06128533184528351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,balanced,0.07181333502133687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,balanced,0.07906133433183034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,balanced,0.09503466884295146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,balanced,0.1671253244082133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,balanced,0.2053226629892985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,balanced,0.2877333362897237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,balanced,0.36004265149434406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,balanced,0.11161067088445027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,balanced,0.17312000195185342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,balanced,0.1739413340886434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,balanced,0.17643199364344278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,balanced,0.17710934082667032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,balanced,0.17634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,balanced,0.17657599846522012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,balanced,0.18020800749460855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,balanced,0.18102399508158365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,balanced,0.18130133549372354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,balanced,0.18871466318766275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,balanced,0.18888533115386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,balanced,0.1925493280092875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,balanced,0.2016106645266215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,balanced,0.20845866203308105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,balanced,0.22430932521820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,balanced,0.23959465821584067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,balanced,0.2861226598421733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,balanced,0.31722132364908856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,balanced,0.39523200194040936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,balanced,0.45027732849121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,balanced,0.6117386817932129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,balanced,0.7332586447397867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,balanced,0.05666666726271311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,balanced,0.058389330903689064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,balanced,0.05973866581916809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,balanced,0.06097066899140676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,balanced,0.06553600231806438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,balanced,0.06605333089828491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,balanced,0.0798880010843277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,balanced,0.08014399806658427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,balanced,0.1032533347606659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,balanced,0.11389866471290588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,balanced,0.14270400007565817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,balanced,0.16963199774424234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,balanced,0.22621333599090576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,balanced,0.2831520040829976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,balanced,0.3978773355484009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,balanced,0.5014986594518026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,balanced,0.7325173219045004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,balanced,0.9504906336466471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.2,0.07150080204010009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.2,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.2,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.2,0.07939839959144593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.2,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.2,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.2,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.2,0.09825279712677001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.2,0.10657919645309448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.2,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.2,0.13552639484405518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.2,0.17065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.2,0.18403840065002441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.2,0.2527359962463379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.2,0.2639616012573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.18875520229339598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.23299839496612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.2,0.3635967969894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.25830399990081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.2,0.43727359771728513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.40177278518676757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.3907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.39070079326629636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.395577597618103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.41466240882873534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.4263040065765381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.44817280769348145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.4455808162689209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.4685823917388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.5011392116546631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.5048704147338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,power_law_1.01,0.5368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,power_law_1.01,0.531987190246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,power_law_1.01,0.6037439823150634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.2,0.6527872085571289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,power_law_1.01,0.6635136127471923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.2,0.7635392189025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,power_law_1.2,0.07346559762954712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,power_law_1.01,0.7999680042266846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,power_law_1.2,0.0736191987991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,power_law_1.01,0.9076031684875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,power_law_1.2,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,power_law_1.2,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,power_law_1.2,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,power_law_1.2,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,power_law_1.2,0.09973120093345642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,power_law_1.2,0.09858559966087341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,power_law_1.2,0.10194560289382934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,power_law_1.2,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,power_law_1.2,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,power_law_1.2,0.1103935956954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,power_law_1.2,0.11755520105361938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,power_law_1.01,1.2042880058288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,power_law_1.2,0.12499200105667115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,power_law_1.2,0.13493759632110597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,power_law_1.2,0.14903680086135865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.2,0.16184959411621094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.2,0.19129600524902343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.2,0.22366719245910643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.2,0.326911997795105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.2,0.4153855800628662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.2,0.5540480136871337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.2,0.699398422241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.2,1.1232831954956055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,power_law_1.01,1.7111040115356446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,power_law_1.01,1.705254364013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.2,1.5883328437805175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,power_law_1.01,3.080121612548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,power_law_1.01,3.849068832397461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,power_law_1.01,5.07531509399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,power_law_1.01,7.168300628662109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,power_law_1.01,0.09530879855155945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,power_law_1.01,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,power_law_1.01,0.23987839221954346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,power_law_1.01,0.23756160736083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,power_law_1.01,0.3468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,power_law_1.01,0.41948800086975097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,power_law_1.01,0.47052798271179197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,power_law_1.01,0.5177728176116944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,power_law_1.01,0.5306560039520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,power_law_1.01,0.5401535987854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,power_law_1.01,0.5577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,power_law_1.01,0.5425280094146728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,power_law_1.01,0.5287551879882812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,power_law_1.01,0.5481088161468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,power_law_1.01,0.5834688186645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,power_law_1.01,0.5538815975189209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,power_law_1.01,0.622489595413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,power_law_1.01,0.637331199645996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,power_law_1.01,0.6977471828460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,power_law_1.01,0.7892223834991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,power_law_1.01,0.8131008148193359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,power_law_1.01,0.9691519737243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,power_law_1.01,1.0716032028198241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,power_law_1.01,1.3958271980285644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,power_law_1.01,1.4522239685058593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,power_law_1.01,1.9503936767578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,power_law_1.01,2.3131071090698243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,power_law_1.2,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,power_law_1.2,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,power_law_1.2,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,power_law_1.2,0.10815999507904053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,power_law_1.2,0.10230400562286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,power_law_1.2,0.09991679787635803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,power_law_1.2,0.10664960145950317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,power_law_1.2,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,power_law_1.2,0.11018240451812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,power_law_1.2,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,power_law_1.2,0.11006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,power_law_1.2,0.1020095944404602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,power_law_1.2,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,power_law_1.2,0.11705600023269654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,power_law_1.2,0.12533119916915894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,power_law_1.2,0.14306559562683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,power_law_1.2,0.1493631958961487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,power_law_1.2,0.1821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,power_law_1.2,0.20833919048309327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,power_law_1.2,0.2703488111495972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,power_law_1.2,0.33997440338134766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,power_law_1.2,0.453004789352417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,power_law_1.2,0.5723072052001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,power_law_1.2,0.8229056358337402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,power_law_1.2,1.107046413421631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.09149439930915833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.15160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.20451838970184327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.26828799247741697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.344812798500061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.43325438499450686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.44216318130493165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.44983677864074706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.42945280075073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.4668032169342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.4770048141479492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.47837438583374026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.4974912166595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.49918718338012696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.524563217163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.5647168159484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.5832767963409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.612499189376831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,0.6938623905181884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,0.7637375831604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.9060544013977051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.0577664375305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.336070442199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.6693952560424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,2.2614784240722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,2.753561592102051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,power_law_1.01,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,power_law_1.01,0.15196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,power_law_1.01,0.25007359981536864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,power_law_1.01,0.4390528202056885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,power_law_1.01,0.7448895931243896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,power_law_1.01,1.0673279762268066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,power_law_1.01,1.2371520042419433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,power_law_1.01,1.593900775909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,power_law_1.01,1.690719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,power_law_1.01,1.7778303146362304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,power_law_1.01,1.8004480361938477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,power_law_1.01,1.8693824768066407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,power_law_1.01,1.9044607162475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,power_law_1.01,1.9762559890747071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,power_law_1.01,2.0622016906738283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,power_law_1.01,2.048633575439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,power_law_1.01,2.146342468261719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,power_law_1.01,2.258502388000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,power_law_1.01,2.3952831268310546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,power_law_1.01,2.6040767669677733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,power_law_1.01,2.7655744552612305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,power_law_1.01,3.1797760009765623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,power_law_1.01,3.361644744873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,balanced,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,balanced,0.08895466725031535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,balanced,0.08789333701133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,balanced,0.08834133545557658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,balanced,0.08809066812197368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,balanced,0.09002133210500081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,balanced,0.08869866530100505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,balanced,0.08833600083986919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,balanced,0.09756799538930257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,balanced,0.09484799702962239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,balanced,0.10126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,balanced,0.10771200060844421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,balanced,0.11315733194351196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,balanced,0.1272213359673818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,balanced,0.14156267046928406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,balanced,0.1698240041732788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,balanced,0.19484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,balanced,0.2469173272450765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,balanced,0.3076266646385193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,balanced,0.4100586573282878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,balanced,0.5192853212356567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,power_law_1.01,4.091379165649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,power_law_1.01,4.426201629638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,power_law_1.01,5.591462326049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,power_law_1.01,6.686688232421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,balanced,0.04818133513132731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,balanced,0.07037333150704701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,balanced,0.10337066650390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,balanced,0.15542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,balanced,0.2568533420562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,balanced,0.3088159958521525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,balanced,0.30905065933863324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,balanced,0.3120906750361125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,balanced,0.31383466720581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,balanced,0.31489066282908124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,balanced,0.31410133838653564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,balanced,0.3165653347969055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,balanced,0.3168479998906453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,balanced,0.3200959960619609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,balanced,0.32711466153462726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,balanced,0.3307253321011861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,balanced,0.3356106678644816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,balanced,0.35689600308736164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,balanced,0.35893865426381427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,balanced,0.43062933286031085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,balanced,0.40594665209452313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,balanced,0.5721813440322876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,balanced,0.5160213311513265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,balanced,0.722762664159139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,balanced,0.7483466466267904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,balanced,1.026911973953247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,balanced,1.3295520146687825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.10165120363235473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.14240000247955323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.150764799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.15804799795150756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.1612223982810974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.16711039543151857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.17132799625396727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.1743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.18252160549163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.18625279664993286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.19351040124893187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.2025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.21521279811859131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.2437824010848999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.27128961086273196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.3327872037887573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.37180800437927247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.4705344200134277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,0.5634943962097168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,0.7602943897247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,0.952940845489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,power_law_1.01,0.06113920211791992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,power_law_1.01,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,1.3382783889770509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,1.7329599380493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,power_law_1.01,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,power_law_1.01,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,power_law_1.01,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,power_law_1.01,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,balanced,0.06407466530799866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,balanced,0.07671999931335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,balanced,0.09488532940546672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,power_law_1.01,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,balanced,0.09723732868830363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,balanced,0.09920000036557515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,balanced,0.09988799691200256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,balanced,0.10087466239929199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,balanced,0.10280000170071919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,balanced,0.10451199611028035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,balanced,0.10325866937637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,power_law_1.01,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,balanced,0.10656533638636272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,balanced,0.11357866724332173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,balanced,0.11553600430488586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,balanced,0.11952533324559529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,balanced,0.1330560048421224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,balanced,0.14511467019716898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,power_law_1.01,0.08622080087661743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,balanced,0.17125866810480753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,balanced,0.19548799594243368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,balanced,0.24105066061019897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,balanced,0.28861866394678753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,balanced,0.41859201590220135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,balanced,0.5267146825790405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,balanced,0.7606240113576254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,balanced,0.9919892946879069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,power_law_1.01,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,power_law_1.01,0.1360703945159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,power_law_1.01,0.19376640319824218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,power_law_1.01,0.24693760871887208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,power_law_1.01,0.35985279083251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,power_law_1.01,0.439737606048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,power_law_1.2,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,power_law_1.2,0.05964159965515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,power_law_1.2,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,power_law_1.2,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,power_law_1.2,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,power_law_1.2,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,power_law_1.2,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,power_law_1.2,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,power_law_1.2,0.09930880069732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,power_law_1.2,0.12511359453201293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,power_law_1.2,0.1447167992591858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,power_law_1.2,0.19459199905395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,power_law_1.2,0.23613440990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,power_law_1.2,0.33739519119262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,power_law_1.2,0.4494272232055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,power_law_1.2,0.6502143859863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,power_law_1.2,0.8538944244384765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,balanced,0.030389333764712017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,balanced,0.029274667302767437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,balanced,0.029194665451844532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,balanced,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,balanced,0.03472000112136205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,balanced,0.03402666747570038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,balanced,0.06211199859778086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,balanced,0.07233066856861115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,balanced,0.09864532947540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,balanced,0.1256480018297831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,balanced,0.16922134160995483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,balanced,0.20989867051442465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.2,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.2,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.2,0.1723199963569641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.2,0.1835263967514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.2,0.19242240190505983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.2,0.19736319780349731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.2,0.19968639612197875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.2,0.20924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.2,0.21496961116790772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.2,0.21288959980010985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.2,0.22368640899658204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.2,0.23945600986480714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.2,0.25189120769500734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.2,0.24417920112609864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.2,0.2780735969543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.2,0.27633280754089357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.2,0.31582720279693605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.2,0.34894719123840334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.2,0.41521282196044923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.2,0.4667520046234131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,power_law_1.01,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,power_law_1.01,0.028512001037597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,power_law_1.01,0.02858879864215851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.2,0.5913983821868897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.2,0.7171904087066651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.2,0.9699968338012696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,power_law_1.01,0.04182400107383728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,power_law_1.01,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.01,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.01,0.11572480201721191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.01,0.14945919513702394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.01,0.19141119718551636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.01,0.237990403175354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.2,1.209228801727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.01,0.36405119895935056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.01,0.5748223781585693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.01,0.7565311908721923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.097324800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.13850239515304566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.17539199590682983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.2482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.3294464111328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.46649599075317383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.48716158866882325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.5199488162994385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.5069183826446533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.5323840141296386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.5245503902435302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.556332778930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.5478591918945312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.5540671825408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.5847104072570801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.5896704196929932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.5979328155517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.6752831935882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,power_law_1.01,0.6888127803802491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,power_law_1.01,0.8400896072387696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,power_law_1.01,0.9592960357666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,power_law_1.01,1.1814784049987792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,power_law_1.01,1.1364992141723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,power_law_1.01,1.3904191970825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,power_law_1.01,1.818124771118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,power_law_1.01,2.43253116607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.01,1.211353588104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,power_law_1.01,2.7948671340942384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,balanced,0.08237333099047343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,balanced,0.08953600128491719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,balanced,0.11024533708890279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,balanced,0.14611732959747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,balanced,0.2176106572151184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,balanced,0.21849600474039713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,balanced,0.22162665923436484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,balanced,0.21847466627756754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,balanced,0.22193600734074911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,balanced,0.22220800320307413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,balanced,0.223855992158254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,balanced,0.22723732391993204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,balanced,0.22758400440216064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,balanced,0.22986666361490884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,4,balanced,0.2312800089518229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,4,balanced,0.23492266734441122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,4,balanced,0.24230400721232095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,4,balanced,0.2532586654027303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,4,balanced,0.27582399050394696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,4,balanced,0.29149866104125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,4,balanced,0.32867199182510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,4,balanced,0.3695840040842692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,4,balanced,0.5117813348770142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,4,balanced,0.6004746754964193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,4,balanced,0.8433226744333903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,4,balanced,1.1028587023417156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,power_law_1.01,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,power_law_1.01,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,power_law_1.01,0.04645760059356689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,power_law_1.01,0.07460479736328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,power_law_1.01,0.0917568027973175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,power_law_1.01,0.11840640306472779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,power_law_1.01,0.14215680360794067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,power_law_1.01,0.20583040714263917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,power_law_1.01,0.3102911949157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,power_law_1.01,0.41194877624511717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,power_law_1.01,0.4701568126678467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,balanced,0.044250667095184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,balanced,0.04818133513132731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,balanced,0.056746666630109154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,balanced,0.06171200176080068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,balanced,0.0688266654809316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,balanced,0.0720000018676122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,balanced,0.08076266447703044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,balanced,0.09353066484133403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,balanced,0.10100799798965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,balanced,0.11949867010116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,balanced,0.14502400159835815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,balanced,0.1812959909439087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,balanced,0.21158399184544882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,balanced,0.28886934121449787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,balanced,0.36158398787180585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,balanced,0.5232053200403849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,balanced,0.6529599825541178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.11943039894104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.16008960008621215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.19207040071487427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.26623361110687255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.34431359767913816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.5003456115722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,balanced,0.031930667658646904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.6383552074432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,balanced,0.0761706680059433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,balanced,0.12027733524640401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,balanced,0.1195093293984731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,balanced,0.12029332915941875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,balanced,0.12300266822179158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,balanced,0.1206773320833842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,balanced,0.12212799986203511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,balanced,0.12290133039156596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,balanced,0.12370133399963379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,balanced,0.12618666887283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,balanced,0.126202662785848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,balanced,0.13090133666992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,balanced,0.1341600020726522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,balanced,0.1360319952170054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,balanced,0.14101866881052652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,balanced,0.1467413306236267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,balanced,0.17070933183034262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,balanced,0.17620799938837686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,balanced,0.2184106707572937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,balanced,0.24136000871658325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,balanced,0.32550932963689166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,balanced,0.37162665526072186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,balanced,0.49659732977549237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,balanced,0.6369440158208212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.06485120058059693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.06446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.07452800273895263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.09264640212059021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.10829440355300904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.11929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.14792959690093993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.17605760097503662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.22202880382537843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.2830143928527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,power_law_1.01,0.37660799026489256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,power_law_1.01,0.5583871841430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,power_law_1.01,0.04449920058250427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,power_law_1.01,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,power_law_1.01,0.07742080092430115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,power_law_1.01,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,power_law_1.01,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,power_law_1.01,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,power_law_1.01,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,power_law_1.01,0.10892159938812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,power_law_1.01,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,power_law_1.01,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,power_law_1.01,0.16378240585327147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,power_law_1.01,0.20049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,power_law_1.01,0.2622783899307251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,power_law_1.01,0.3331648111343384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,power_law_1.01,0.44929919242858884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,power_law_1.01,0.5598527908325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,power_law_1.01,0.8068032264709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,power_law_1.01,1.0006912231445313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.10741759538650512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.1723199963569641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.22984960079193115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.24024319648742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.244486403465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.2524928092956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.2636800050735474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.2740479946136475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.28705921173095705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.2929984092712402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.2966079950332642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.31472001075744627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.3303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.3813375949859619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.4154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5026624202728271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.5785984039306641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.7331647872924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.2056832313537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.508403205871582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.1025920867919923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.712960052490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.01,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.01,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.01,0.0423552006483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.01,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.01,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.01,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.01,0.10353920459747315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.01,0.15856640338897704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.01,0.16910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,balanced,0.03391999999682108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,balanced,0.03290133426586787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,balanced,0.035162667433420815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,balanced,0.03708266715208689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,balanced,0.05751466751098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,balanced,0.07513600091139476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,balanced,0.08900266885757446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,balanced,0.11545600493748982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,balanced,0.1349493364493052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,balanced,0.18050666650136313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,balanced,0.22542399168014526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,balanced,0.31309332450230914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,balanced,0.3981599807739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,power_law_1.01,0.045459198951721194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,power_law_1.01,0.08425599932670594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,power_law_1.01,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,power_law_1.01,0.1223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,power_law_1.01,0.12339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,power_law_1.01,0.12928639650344848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,power_law_1.01,0.13860479593276978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,power_law_1.01,0.14140160083770753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,power_law_1.01,0.15105279684066772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,power_law_1.01,0.16617599725723267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,power_law_1.01,0.1740224003791809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,power_law_1.01,0.17354240417480468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,power_law_1.01,0.20371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,power_law_1.01,0.22965760231018068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,power_law_1.01,0.27922561168670657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,power_law_1.01,0.3319936037063599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,power_law_1.01,0.4370880126953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,power_law_1.01,0.5475456237792968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,power_law_1.01,0.7442368030548095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,power_law_1.01,0.9574080467224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,power_law_1.01,1.36114559173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,power_law_1.01,1.7243648529052735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,power_law_1.01,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,power_law_1.01,0.05868800282478333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,power_law_1.01,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,power_law_1.01,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,power_law_1.01,0.05847679972648621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,power_law_1.01,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,power_law_1.01,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,power_law_1.01,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,power_law_1.01,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,power_law_1.01,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,power_law_1.01,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,power_law_1.01,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,power_law_1.01,0.08275200128555298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,power_law_1.01,0.0821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,power_law_1.01,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,power_law_1.01,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,power_law_1.01,0.08385279774665833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,power_law_1.01,0.1229632019996643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,power_law_1.01,0.14320000410079955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,power_law_1.01,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,power_law_1.01,0.18196480274200438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,power_law_1.01,0.22308480739593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,power_law_1.01,0.306112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,power_law_1.01,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,power_law_1.01,0.3931328058242798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,power_law_1.01,0.545081615447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,power_law_1.01,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,power_law_1.01,0.11415040493011475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,power_law_1.01,0.12225279808044434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,power_law_1.01,0.14908159971237184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,power_law_1.01,0.17221759557723998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,power_law_1.01,0.22197120189666747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,power_law_1.01,0.2739903926849365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,power_law_1.01,0.721401596069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,power_law_1.01,0.3720896005630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,power_law_1.01,0.4632512092590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,power_law_1.01,0.6799232006072998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,power_law_1.01,0.848697566986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.1089792013168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.11893759965896607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.13687039613723756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.1549631953239441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.18644479513168336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.22739200592041015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.2945535898208618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.3693631887435913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.5284671783447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.6501567840576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.931884765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.2120960235595704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.07970560193061829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.107315194606781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.14853760004043579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.17679359912872314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.23045759201049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.2953088045120239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.40623998641967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.5215807914733886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,0.7533504009246826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,balanced,0.05681066711743673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,balanced,0.10973866780598958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,balanced,0.17237865924835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,balanced,0.3004106680552165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,balanced,0.5540106693903605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,balanced,0.5541333357493082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,balanced,0.5548426707585653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,balanced,0.5580480098724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,balanced,0.5599786837895712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,balanced,0.5594346523284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,balanced,0.5707626740137736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,balanced,0.5696320136388143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,balanced,0.576042652130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,balanced,0.583898663520813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,balanced,0.5861440102259318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,balanced,0.5997920036315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,balanced,0.6248106559117635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,balanced,0.6467359860738119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,balanced,0.6993280251820883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,balanced,0.7521226406097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,balanced,0.8215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,balanced,0.9046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,balanced,1.1383573214213054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,balanced,1.2825173536936443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,balanced,1.789141337076823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,balanced,2.1424320538838706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,0.9912832260131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,power_law_1.01,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,power_law_1.01,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,power_law_1.01,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,power_law_1.01,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,power_law_1.01,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,power_law_1.01,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,power_law_1.01,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,power_law_1.01,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,power_law_1.01,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,power_law_1.01,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.01,0.09186559915542603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.01,0.11112960577011108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.01,0.13861759901046752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.01,0.25484800338745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.01,0.3177151918411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.01,0.4646848201751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.01,0.5799359798431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,power_law_1.01,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,power_law_1.01,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,power_law_1.01,0.04378879964351654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,power_law_1.01,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,power_law_1.01,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,power_law_1.01,0.12890880107879638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,power_law_1.01,0.15971200466156005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,power_law_1.01,0.20541439056396485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,power_law_1.01,0.26759040355682373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,power_law_1.2,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,power_law_1.2,0.09118080139160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,power_law_1.2,0.12040319442749023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,power_law_1.2,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,power_law_1.2,0.19118720293045044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,power_law_1.2,0.22120959758758546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,power_law_1.2,0.2649343967437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,power_law_1.2,0.2777728080749512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,power_law_1.2,0.2830399990081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,power_law_1.2,0.30012800693511965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,power_law_1.2,0.30010879039764404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,power_law_1.2,0.303603196144104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,power_law_1.2,0.32581119537353515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,power_law_1.2,0.34413440227508546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,power_law_1.2,0.3417855978012085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,power_law_1.2,0.36559998989105225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,power_law_1.2,0.39711999893188477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,power_law_1.2,0.43416318893432615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,power_law_1.2,0.5223487854003906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,power_law_1.2,0.58373122215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,power_law_1.2,0.7351359844207763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,power_law_1.2,0.9181887626647949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,power_law_1.2,1.1836031913757323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,power_law_1.2,1.3661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,power_law_1.2,1.7923519134521484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,power_law_1.2,2.437580871582031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.07507200241088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.08483840227127075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.15966720581054689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.21697280406951905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.2427583932876587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,power_law_1.01,0.3627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,power_law_1.01,0.4989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,power_law_1.01,0.6316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,power_law_1.01,0.8559359550476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,power_law_1.2,0.048332801461219786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,power_law_1.2,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,power_law_1.2,0.06221439838409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,power_law_1.2,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,power_law_1.2,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,power_law_1.2,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,power_law_1.2,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,power_law_1.2,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.2,0.11620479822158813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.2,0.14231040477752685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.2,0.17562880516052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.2,0.22358400821685792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.2,0.3471872091293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.2,0.43212161064147947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,balanced,0.07923733194669087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,balanced,0.08052800099054973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,balanced,0.07954666515191396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,balanced,0.08216533561547597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,balanced,0.0839573343594869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,balanced,0.0841919978459676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,balanced,0.08457600076993306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,balanced,0.09306666254997253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,balanced,0.09291733304659526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,balanced,0.09923733274141948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,balanced,0.11125866572062175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,balanced,0.11780800422032674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,balanced,0.13609066605567932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,balanced,0.15057067076365152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,balanced,0.18444265921910605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,balanced,0.23371734221776327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,balanced,0.3075253367424011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,balanced,0.38811198870340985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,balanced,0.5405653317769369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,balanced,0.6832373142242432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,power_law_1.2,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,power_law_1.2,0.04627839922904968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,power_law_1.2,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,power_law_1.2,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,power_law_1.2,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,power_law_1.2,0.055180799961090085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,power_law_1.2,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,power_law_1.2,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,power_law_1.2,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,power_law_1.2,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,power_law_1.2,0.1289023995399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,power_law_1.2,0.1486400008201599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,power_law_1.2,0.17827199697494506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,power_law_1.2,0.2169343948364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,power_law_1.2,0.3025151968002319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,power_law_1.2,0.3877311944961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,power_law_1.2,0.5147776126861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.2,0.6103360176086425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,power_law_1.2,0.7061312198638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.2,0.9019071578979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.2,1.0658687591552733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.08510720133781433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.08028159737586975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.09244160056114196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.116428804397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.1321727991104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.15051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.19217920303344727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.23867518901824952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.31289598941802976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.3860863924026489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.5981696128845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,power_law_1.01,0.07603840231895446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.7394112110137939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,power_law_1.01,0.10498559474945068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,power_law_1.01,0.170905601978302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,power_law_1.01,0.2312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,power_law_1.01,0.29795839786529543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,power_law_1.01,0.4099775791168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,power_law_1.01,0.4283328056335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,power_law_1.01,0.4436351776123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,power_law_1.01,0.4619647979736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,power_law_1.01,0.4666751861572266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,power_law_1.01,0.4838592052459717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,power_law_1.01,0.48426241874694825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,power_law_1.01,0.49898881912231446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,power_law_1.01,0.5142144203186035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,power_law_1.01,0.5291967868804932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,power_law_1.01,0.558400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,power_law_1.01,0.6013311862945556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,power_law_1.01,0.6136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,power_law_1.01,0.6912767887115479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,power_law_1.01,0.7250432014465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,power_law_1.01,0.8181056022644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,power_law_1.01,0.934540843963623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,power_law_1.01,1.1770560264587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,power_law_1.01,1.416710376739502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,power_law_1.01,1.8827007293701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,power_law_1.01,2.3708927154541017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.1347648024559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.16116479635238648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.18917759656906127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.24737279415130614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.2948992013931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.40446720123291013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.5121088027954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.72707200050354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.9286463737487793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,balanced,0.12717333436012268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,balanced,0.2126880089441935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,balanced,0.21426665782928467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,balanced,0.2145706613858541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,balanced,0.2130026618639628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,balanced,0.21480000019073486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,balanced,0.21401600042978922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,balanced,0.2193173368771871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,balanced,0.22049599885940552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,balanced,0.22126400470733643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,balanced,0.22269866863886514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,balanced,0.22362132867177328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,balanced,0.2403786579767863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,balanced,0.2510773340861003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,balanced,0.25473066171010333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,balanced,0.26341332991917926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,balanced,0.31618134180704754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,balanced,0.32870399951934814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,balanced,0.4043733278910319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,balanced,0.47972798347473145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,balanced,0.6399146715799967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,balanced,0.8021706740061442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,balanced,1.134874661763509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,balanced,1.5108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,balanced,2.2168374061584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,balanced,2.8936373392740884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,2,balanced,0.1111840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,2,balanced,0.11994133392969768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,2,balanced,0.13272533814112344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,2,balanced,0.16768532991409302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,2,balanced,0.239519993464152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,2,balanced,0.3838293155034383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,2,balanced,0.38553067048390705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,2,balanced,0.38909868399302167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,2,balanced,0.3877013524373372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,2,balanced,0.39107199509938556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,2,balanced,0.3917066653569539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,2,balanced,0.39476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,2,balanced,0.3970880111058553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,2,balanced,0.40087465445200604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,2,balanced,0.40067732334136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,2,balanced,0.40267733732859295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,2,balanced,0.41038934389750165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,2,balanced,0.42580266793568927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,2,balanced,0.43858667214711505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,2,balanced,0.4705333312352498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,2,balanced,0.5016479889551798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,2,balanced,0.5632853507995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,2,balanced,0.6345813274383545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,2,balanced,0.8757493495941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,2,balanced,1.0146133104960124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,2,balanced,1.4214240709940593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,2,balanced,1.8691786130269368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.11653120517730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.15811840295791627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.15022079944610595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.15441919565200807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.144595205783844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.1357632040977478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.12441600561141967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.15054719448089598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.1470080018043518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.14881279468536376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.14565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.14049279689788818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.15267200469970704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.16584960222244263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.15711359977722167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.16224000453948975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.1765247941017151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.18632960319519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.21636478900909423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.2300352096557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.2920383930206299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.36796159744262696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.4749120235443115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,power_law_1.2,0.5970943927764892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,power_law_1.2,0.7668416023254394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,power_law_1.2,1.086124801635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,balanced,0.04673066735267639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.01,0.12952959537506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,balanced,0.04614399870236715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,balanced,0.0543093333641688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.01,0.1714560031890869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.01,0.18702720403671264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.01,0.2617408037185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.01,0.3324160099029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,balanced,0.05645333230495453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,balanced,0.058730666836102806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,balanced,0.057317331433296204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,balanced,0.05981333553791046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,balanced,0.06221333146095276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.01,0.4898496150970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,balanced,0.07040533423423767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,balanced,0.07424533367156982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,balanced,0.08686400453249614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,balanced,0.09887466828028361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.01,0.5081471920013427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,balanced,0.12020799517631531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,balanced,0.135343998670578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,balanced,0.1723733345667521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,balanced,0.20468266805013022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.01,0.5275008201599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,balanced,0.2784693241119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,balanced,0.343450665473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.01,0.5283199787139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.01,0.551852798461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.01,0.5623487949371337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.01,0.5699007987976075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.01,0.5818367958068847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.01,0.5901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.01,0.6113408088684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.01,0.6295936107635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.01,0.6545087814331054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.01,0.7273151874542236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.01,0.8026432037353516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.01,0.9784768104553223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.01,1.0530431747436524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.01,1.3210240364074708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.01,1.5932607650756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.01,2.1194623947143554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.01,2.671731185913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.01,3.8704769134521486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.01,4.89312629699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.2,0.027116799354553224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.2,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.2,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.2,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.2,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.2,0.12646399736404418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.2,0.1973312020301819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.2,0.276307201385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.2,0.31512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,balanced,0.03186133255561193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,balanced,0.059434667229652405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,balanced,0.06171200176080068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,balanced,0.0636053333679835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,balanced,0.06886399785677592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,balanced,0.06799466907978058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,balanced,0.07066666583220164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,balanced,0.0745600014925003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,balanced,0.08721066514650981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,balanced,0.09779733419418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,balanced,0.11994666854540507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,balanced,0.13902933398882547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,balanced,0.179584006468455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,balanced,0.22380799055099487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,balanced,0.3041013280550639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,balanced,0.3718986511230469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,balanced,0.0783733328183492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,balanced,0.09797333677609761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,balanced,0.11142933368682861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,balanced,0.11351999640464783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,balanced,0.11216533184051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,balanced,0.1125866671403249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,balanced,0.11416000127792358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,balanced,0.1163200040658315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,balanced,0.11827199657758077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,balanced,0.1200373371442159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,balanced,0.12130133310953777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,balanced,0.1270240048567454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,balanced,0.13145066301027933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,balanced,0.1357866624991099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,balanced,0.150218665599823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,balanced,0.16102400422096252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,balanced,0.21196800470352173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,balanced,0.2267413338025411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,balanced,0.3198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,balanced,0.3086186647415161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,balanced,0.4413226842880249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,balanced,0.5164159933725992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,balanced,0.7289653619130453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,balanced,0.8857119878133138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,balanced,0.03387200087308884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,balanced,0.0444213350613912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,balanced,0.04398400088151296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.05489066739877065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.0804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.10327999790509541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.12377599875132243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.15944533546765646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.1866933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,power_law_1.01,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,power_law_1.01,0.08721280097961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,power_law_1.01,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,power_law_1.01,0.1192512035369873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,power_law_1.01,0.13146239519119263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,power_law_1.01,0.13498879671096803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,power_law_1.01,0.13735040426254272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,power_law_1.01,0.13912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,power_law_1.01,0.14279680252075194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,power_law_1.01,0.14636160135269166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,power_law_1.01,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,power_law_1.01,0.1480064034461975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,power_law_1.01,0.15061119794845582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,power_law_1.01,0.15393279790878295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,power_law_1.01,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,power_law_1.01,0.16716159582138063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.01,0.18148479461669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.01,0.20897281169891357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.01,0.24005119800567626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.01,0.28803200721740724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.01,0.34419200420379636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.01,0.47446398735046386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.01,0.6197247982025147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.01,0.8536959648132324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.01,1.1293567657470702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.08306559920310974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.09205759763717651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.10680320262908935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.1449023962020874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.14729599952697753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.16398719549179078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.18121600151062012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.13303040266036986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.171833598613739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.1760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.2388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.2714495897293091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,power_law_1.01,0.3521343946456909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,power_law_1.01,0.39765760898590086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,power_law_1.01,0.5477503776550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,power_law_1.01,0.8128447532653809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,power_law_1.01,1.162764835357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,power_law_1.01,1.4640255928039552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,power_law_1.01,2.0604480743408202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,power_law_1.01,2.8971519470214844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,balanced,0.06961599985758464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,balanced,0.09803733229637146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,balanced,0.146997332572937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,balanced,0.14874133467674255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,balanced,0.15077867110570273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,balanced,0.152319997549057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,balanced,0.15108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,balanced,0.15250666936238608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,balanced,0.15212266643842062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,balanced,0.15331733226776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,balanced,0.15547733505566916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,balanced,0.16214399536450705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,balanced,0.16156267126401266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,balanced,0.16767466068267822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,balanced,0.1753386656443278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,balanced,0.1798080007235209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,balanced,0.19853333632151285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,balanced,0.2233546574910482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,balanced,0.2561226685841878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,balanced,0.28492265939712524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,balanced,0.34857598940531415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,balanced,0.408517320950826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,balanced,0.5751359860102335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,balanced,0.6948426564534506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,balanced,0.03807999938726425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,balanced,0.04058133314053217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,balanced,0.07473066449165344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,balanced,0.09107200304667155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,balanced,0.10065066814422607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,balanced,0.13263466954231262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,balanced,0.14577600359916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,balanced,0.20166399081548056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,balanced,0.23996265729268393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,balanced,0.3450293143590291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,balanced,0.4432266553243001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,balanced,0.637440005938212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,balanced,0.8222080071767172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,power_law_1.01,0.029734399914741517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,power_law_1.01,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,power_law_1.01,0.029216000437736513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,power_law_1.01,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,power_law_1.01,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,power_law_1.01,0.07280640006065368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.10454399585723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.11507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,power_law_1.2,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,power_law_1.01,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.1251904010772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,power_law_1.2,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,power_law_1.2,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.12721279859542847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,power_law_1.2,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.13043839931488038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.13252480030059816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.13923200368881225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,power_law_1.01,0.11003520488739013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,power_law_1.2,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,power_law_1.2,0.04993920028209686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.14419840574264525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.14862719774246216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.15130239725112915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.1738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,power_law_1.2,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.1640447974205017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,power_law_1.2,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,power_law_1.2,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.1801792025566101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,power_law_1.2,0.0896127998828888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.1836159944534302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,power_law_1.2,0.09716479778289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.21847679615020751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,power_law_1.2,0.11654399633407593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.24443519115447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,power_law_1.2,0.15057920217514037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.2961663961410522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,power_law_1.2,0.18653440475463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.3526400089263916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,power_law_1.2,0.23429760932922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,power_law_1.01,0.4198592185974121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,power_law_1.2,0.3280832052230835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,power_law_1.01,0.510214376449585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,power_law_1.01,0.6922175884246826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,power_law_1.01,0.9588864326477051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,power_law_1.2,0.42911357879638673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,power_law_1.01,0.17817599773406984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,power_law_1.2,0.5907711982727051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,power_law_1.01,0.2153536081314087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,power_law_1.2,0.7875199794769288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,power_law_1.01,0.3007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.06229760050773621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.11780480146408082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.16976640224456788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.18186240196228026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.18968960046768188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.18684159517288207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.18992639780044557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.20019199848175048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.20488319396972657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.20133121013641359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.21480960845947267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.2233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.2277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.24209280014038087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.27251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.2613055944442749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.3060992002487183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.3351615905761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.39673600196838377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.01,0.4564864158630371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.01,0.5672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.01,0.6610623836517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.01,0.8982848167419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.01,1.125657558441162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,power_law_1.01,0.37900800704956056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,power_law_1.01,0.5070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,power_law_1.01,0.6993599891662597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,power_law_1.01,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,power_law_1.01,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,power_law_1.01,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,power_law_1.01,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,power_law_1.01,0.06336640119552613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,power_law_1.01,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.01,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.01,0.09763839840888977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.01,0.10520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.01,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.01,0.16131839752197266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.01,0.22659199237823485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.01,0.2649087905883789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.01,0.3808896064758301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.01,0.48779520988464353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,balanced,0.05682666599750519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,balanced,0.0577706644932429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,balanced,0.08077333370844524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,balanced,0.11597333351771037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,balanced,0.11734400192896526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,balanced,0.12129599849383037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,balanced,0.12171733379364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,balanced,0.12160000205039978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,balanced,0.12389333049456279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,balanced,0.1251253286997477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,balanced,0.12622400124867758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,balanced,0.1279306709766388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,balanced,0.13571199774742126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,balanced,0.13683733344078064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,balanced,0.13672000169754028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,balanced,0.15403733650843301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.16172266999880472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.18543465932210287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.20807466904322305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.2564586599667867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.3049599925676982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.43320000171661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.5409119923909506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.7789013385772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,balanced,1.0086987018585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.08615679740905761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.11873279809951783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.14083839654922486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.18568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.22864000797271727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.30899200439453123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,power_law_1.2,0.35041279792785646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.11409280300140381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.11869440078735352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.12250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.13050880432128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.15173120498657228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.17242239713668822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.16971520185470582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.1983423948287964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.23600640296936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.2896320104598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.3324415922164917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.4288320064544678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.01,0.5244736194610595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.01,0.7264895915985108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.01,0.9373632431030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,power_law_1.2,0.5705408096313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,power_law_1.2,0.765177583694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,power_law_1.2,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,power_law_1.2,0.06391040086746216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,power_law_1.2,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,power_law_1.2,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,power_law_1.2,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,power_law_1.2,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,power_law_1.2,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,power_law_1.2,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,power_law_1.2,0.09256319999694824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,power_law_1.2,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,power_law_1.2,0.1001855969429016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,power_law_1.2,0.11266560554504394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.2,0.1256384015083313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.2,0.15137920379638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.2,0.17564799785614013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.2,0.2417151927947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.2,0.3519615888595581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.2,0.5133632183074951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.2,0.6509696006774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.2,0.9480447769165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.2,1.4005120277404786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,power_law_1.01,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,power_law_1.01,0.05968639850616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,power_law_1.01,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,power_law_1.01,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,power_law_1.01,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,power_law_1.01,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,power_law_1.01,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,power_law_1.01,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,power_law_1.01,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,power_law_1.01,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,power_law_1.01,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,power_law_1.01,0.09284480214118958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,power_law_1.01,0.10894080400466918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,power_law_1.01,0.13306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,power_law_1.01,0.15823999643325806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,power_law_1.01,0.20208640098571778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,power_law_1.01,0.24970240592956544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,power_law_1.01,0.3407936096191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,power_law_1.01,0.4329343795776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,power_law_1.01,0.6190080165863037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,power_law_1.01,0.7946303844451904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,power_law_1.01,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,power_law_1.01,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,power_law_1.01,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,power_law_1.01,0.10994559526443481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,power_law_1.01,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,power_law_1.01,0.14899840354919433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,power_law_1.01,0.15705599784851074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,power_law_1.01,0.15605759620666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.2,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,power_law_1.01,0.16922880411148072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.2,0.025887998938560485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.2,0.027296000719070436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.2,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.2,0.028095999360084535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.2,0.02908160090446472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.2,0.028595200181007384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.2,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.2,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,power_law_1.01,0.16577279567718506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.2,0.030239999294281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.2,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,power_law_1.01,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,power_law_1.01,0.1751423954963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,power_law_1.01,0.17885440587997437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,power_law_1.01,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,power_law_1.01,0.19245439767837524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,power_law_1.01,0.20805759429931642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,power_law_1.01,0.23454720973968507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.2,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,power_law_1.01,0.2431936025619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.2,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,power_law_1.01,0.30353920459747313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.2,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,power_law_1.01,0.32425599098205565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,power_law_1.01,0.4063744068145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,power_law_1.01,0.47881598472595216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,power_law_1.01,0.6349503993988037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,power_law_1.01,0.7950975894927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.2,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,power_law_1.01,1.118067169189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,power_law_1.01,1.4439040184020997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.2,0.16568319797515868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.2,0.21918718814849852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,balanced,0.03600533306598663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,balanced,0.03994666785001755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,balanced,0.043280000487963356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,balanced,0.04456000030040741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,balanced,0.045552000403404236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,balanced,0.051701332132021584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,balanced,0.06530666848023732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.08870933453241985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.09773866335550944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.11223999659220378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.16963734229405722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.22618132829666138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.28936533133188885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.406607985496521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.5162986516952515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.2,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.2,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.2,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.2,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.2,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.2,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.2,0.07556480169296265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.2,0.07775359749794006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.2,0.07906559705734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.2,0.08791040182113648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.2,0.08970239758491516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.2,0.09798399806022644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.2,0.11514240503311157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.2,0.12920960187911987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.2,0.16726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.2,0.19544320106506347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.2,0.2550015926361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.2,0.2660608053207397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.2,0.37597439289093015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.2,0.47531518936157224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.2,0.6524288177490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.2,0.8162176132202148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.01,0.08053759932518005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.01,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.01,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.01,0.09785599708557129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.01,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.01,0.10223360061645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.01,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.01,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.01,0.10769280195236205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.01,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.01,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.01,0.13917440176010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.01,0.1481279969215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.01,0.1860159993171692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.01,0.21992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.01,0.2764607906341553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,balanced,0.06806933383146922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,balanced,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,balanced,0.09781866272290547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,balanced,0.09970133503278096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,balanced,0.10154133041699727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.01,0.29178240299224856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,balanced,0.10157333811124165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,balanced,0.10408000151316325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,balanced,0.10737599929173787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,balanced,0.10910933216412862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,balanced,0.11069867014884949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,balanced,0.11572266618410747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.01,0.4032576084136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,balanced,0.12407466769218445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,balanced,0.12798933188120523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,balanced,0.1371893286705017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,balanced,0.15915733575820923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,balanced,0.18184532721837363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,balanced,0.2254719932874044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.01,0.46732158660888673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,balanced,0.2664053241411845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,balanced,0.389850656191508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,balanced,0.47996799151102704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.01,0.6223360061645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,balanced,0.6968586444854736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,balanced,0.8990933100382487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.01,0.7505087852478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,balanced,1.3166613578796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,balanced,1.730837345123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,power_law_1.01,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,power_law_1.01,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,power_law_1.01,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,power_law_1.01,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,power_law_1.01,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,power_law_1.01,0.09225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,power_law_1.01,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,power_law_1.01,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,power_law_1.01,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,power_law_1.01,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,power_law_1.01,0.10949759483337403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,power_law_1.01,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,power_law_1.01,0.12222720384597778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.01,0.14393600225448608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.01,0.1585536003112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.01,0.19107840061187745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.01,0.24456319808959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.01,0.2970880031585693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.01,0.4299136161804199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.01,0.6203008174896241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.01,0.9572159767150878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.01,1.1015040397644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.06347519755363465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.08015360236167908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.08796799778938294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.08513919711112976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.08561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.08389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.10084480047225952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.1093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.1301632046699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.1781056046485901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.19095679521560668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.24398720264434814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.3142591953277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.4004672050476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.5367872238159179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.8000703811645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,0.981049633026123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,power_law_1.01,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,power_law_1.01,0.11472640037536622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,power_law_1.01,0.11824640035629272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,power_law_1.01,0.12254719734191895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,power_law_1.01,0.1255679965019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,power_law_1.01,0.12833280563354493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,power_law_1.01,0.1321727991104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,power_law_1.01,0.1351040005683899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,power_law_1.01,0.13771519660949708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,power_law_1.01,0.1436352014541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,power_law_1.01,0.14812159538269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,power_law_1.01,0.1539903998374939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,power_law_1.01,0.16855679750442504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,power_law_1.01,0.18778239488601683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.01,0.21310720443725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.01,0.26104960441589353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.01,0.3078655958175659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.01,0.40248961448669435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.01,0.5147456169128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.01,0.7045440196990966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.01,0.9225728034973144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.01,1.2988927841186524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.01,1.6545791625976562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,power_law_1.01,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,power_law_1.01,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,power_law_1.01,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,power_law_1.01,0.09512320160865784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,power_law_1.01,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,power_law_1.01,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,power_law_1.01,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,power_law_1.01,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,power_law_1.01,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,power_law_1.01,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,power_law_1.01,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,power_law_1.01,0.1085055947303772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,power_law_1.01,0.11481599807739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,power_law_1.01,0.11743999719619751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,power_law_1.01,0.11963520050048829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,power_law_1.01,0.13754240274429322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,power_law_1.01,0.14658559560775758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,power_law_1.01,0.18094079494476317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,power_law_1.01,0.20587520599365233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,power_law_1.01,0.2552639961242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,power_law_1.01,0.2807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,power_law_1.01,0.3999232053756714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,power_law_1.01,0.504915189743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,power_law_1.01,0.6796927928924561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,power_law_1.01,0.9101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,balanced,0.09347200393676758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,balanced,0.11961066722869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,balanced,0.16294399897257486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,balanced,0.2600053350130717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,balanced,0.4290506839752197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,balanced,0.5913333495457967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,balanced,0.6023679971694946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,balanced,0.6050666570663452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,balanced,0.6077706813812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,balanced,0.6106400092442831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,balanced,0.6191093524297079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,balanced,0.6259946823120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,balanced,0.6297440131505331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,balanced,0.6391253471374512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,balanced,0.6518453359603882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,balanced,0.661242683728536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,balanced,0.6832959651947021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,balanced,0.7315093676249186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,balanced,0.7756213347117106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,balanced,1.0778613090515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,balanced,0.9821013609568278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,balanced,1.3644213676452637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,balanced,1.4433226585388184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,balanced,2.1423999468485513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,balanced,2.403125286102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,balanced,3.7006346384684243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,balanced,4.323941230773926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,power_law_1.2,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,power_law_1.2,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,power_law_1.2,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,power_law_1.2,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,power_law_1.2,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,power_law_1.2,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,power_law_1.2,0.16509439945220947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,power_law_1.2,0.2171328067779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,power_law_1.2,0.2830976009368896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,power_law_1.2,0.39742720127105713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,power_law_1.2,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,power_law_1.2,0.14337919950485228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,power_law_1.2,0.1539520025253296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,power_law_1.2,0.15720959901809692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,power_law_1.2,0.16178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,power_law_1.2,0.1742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,power_law_1.2,0.1869503974914551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,power_law_1.2,0.19315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,power_law_1.2,0.2109055995941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,power_law_1.2,0.23016960620880128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,power_law_1.2,0.24284799098968507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,power_law_1.2,0.24003200531005858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,power_law_1.2,0.29086079597473147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,power_law_1.2,0.33293440341949465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,power_law_1.2,0.4194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,power_law_1.2,0.49272961616516114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,power_law_1.2,0.6624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,power_law_1.2,0.8293888092041015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,power_law_1.2,1.1374527931213378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,power_law_1.2,1.4497407913208007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,power_law_1.2,2.070419120788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,power_law_1.2,2.7220928192138674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,balanced,0.05123733480771383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,balanced,0.05339199801286062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,balanced,0.05525333185990652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,balanced,0.05406400064627329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,balanced,0.057487999399503074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,balanced,0.05745600163936615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,balanced,0.05861866474151611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,balanced,0.06080533564090729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,balanced,0.0678773323694865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,balanced,0.06911466519037883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,balanced,0.07004799942175548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,balanced,0.07965333263079326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,balanced,0.09482133388519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,balanced,0.10150933265686035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,balanced,0.12542933225631714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,balanced,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,balanced,0.196560005346934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,balanced,0.24888533353805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,balanced,0.339957316716512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,balanced,0.4134506781895955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,power_law_1.2,0.5286208152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,balanced,0.025381334125995636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,balanced,0.025146665672461193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,balanced,0.027642667293548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,balanced,0.05106133222579956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,balanced,0.05132266879081726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,balanced,0.05227200190226237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,balanced,0.052906667192777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,balanced,0.059194669127464294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,balanced,0.06312533219655354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,balanced,0.0641599992911021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,balanced,0.06596266726652782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,balanced,0.0779306689898173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,balanced,0.0993226667245229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,balanced,0.1267039974530538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,balanced,0.14873066544532776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,balanced,0.19330666462580362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,balanced,0.233024001121521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,power_law_1.01,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,power_law_1.01,0.05791360139846802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,power_law_1.01,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,power_law_1.01,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,power_law_1.01,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,power_law_1.01,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.01,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.01,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.01,0.12332160472869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.01,0.15363839864730836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.01,0.1817023992538452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.01,0.24685440063476563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.01,0.31290240287780763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.01,0.4416384220123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.01,0.6242496013641358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.08611199855804444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.10027519464492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.11093759536743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.11528960466384888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.1196287989616394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.14086400270462035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.16870399713516235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.20733439922332764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.35437440872192383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.41675519943237305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.5874303817749024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.7499392032623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.0675968170166015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.396998405456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,power_law_1.2,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,power_law_1.2,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,power_law_1.2,0.1348736047744751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,power_law_1.2,0.16167680025100709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,power_law_1.2,0.19425920248031617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,power_law_1.2,0.1996415972709656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,power_law_1.2,0.20683519840240477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,power_law_1.2,0.21673600673675536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,power_law_1.2,0.21226880550384522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,power_law_1.2,0.2267967939376831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,power_law_1.2,0.23717761039733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,power_law_1.2,0.241593599319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,power_law_1.2,0.2388672113418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,power_law_1.2,0.2536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,power_law_1.2,0.2526592016220093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,power_law_1.2,0.2675071954727173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,power_law_1.2,0.31663360595703127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,power_law_1.2,0.3243839979171753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,power_law_1.2,0.39634559154510496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,power_law_1.2,0.4207295894622803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,power_law_1.2,0.545363187789917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,power_law_1.2,0.6135295867919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,power_law_1.2,0.7870207786560058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,power_law_1.2,0.9626432418823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,power_law_1.2,1.3356608390808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,power_law_1.2,1.6942079544067383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,balanced,0.023376000424226124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,balanced,0.025242666403452556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,balanced,0.025600001215934753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,balanced,0.06506133576234181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,balanced,0.05423999826113383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,balanced,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,balanced,0.05774933099746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,balanced,0.06458666423956554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,balanced,0.07072000205516815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,balanced,0.07089066505432129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,balanced,0.06940266489982605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,balanced,0.08391466736793518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,balanced,0.08753599723180135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,balanced,0.11071466406186421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,balanced,0.11199999849001567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,balanced,0.14442132910092673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,balanced,0.16865599155426025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,balanced,0.24041599035263062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,balanced,0.30580800771713257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,balanced,0.4358186721801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,balanced,0.5702986717224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,balanced,0.025098666548728943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,balanced,0.025301332275072735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,balanced,0.027434666951497395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,balanced,0.027189334233601887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,balanced,0.035887998839219414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,balanced,0.0684799998998642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,balanced,0.0776693324247996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,balanced,0.0981280008951823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,balanced,0.11317333579063416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,balanced,0.14824000000953674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,balanced,0.1787946621576945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,balanced,0.24884267648061117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,balanced,0.31379199028015137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,power_law_1.01,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,power_law_1.01,0.02773120105266571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,power_law_1.01,0.02743679881095886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,power_law_1.01,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,power_law_1.01,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,power_law_1.01,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,power_law_1.01,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,power_law_1.01,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,power_law_1.01,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,power_law_1.01,0.05920640230178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,power_law_1.01,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,power_law_1.01,0.11084799766540528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,power_law_1.01,0.1679487943649292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,power_law_1.01,0.19962879419326782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,power_law_1.01,0.29164800643920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,power_law_1.01,0.4421823978424072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.07395200133323669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.10273920297622681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.11756160259246826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.1522304058074951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.1819264054298401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.251692795753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.3240895986557007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.4633471965789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.5987840175628663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,balanced,0.029343999922275543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,balanced,0.031013332307338715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,balanced,0.02937600016593933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,balanced,0.03120533376932144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,balanced,0.031114667654037476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,balanced,0.036303999523321785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,balanced,0.03679466744263967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,balanced,0.07906133433183034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,balanced,0.09320533275604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,power_law_1.01,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,power_law_1.01,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,power_law_1.01,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,power_law_1.01,0.07316480278968811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,power_law_1.01,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,power_law_1.01,0.0779583990573883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,power_law_1.01,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,power_law_1.01,0.09315199851989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,power_law_1.01,0.095769602060318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,power_law_1.01,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,power_law_1.01,0.11016319990158081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,power_law_1.01,0.12944639921188356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,power_law_1.01,0.15342719554901124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,power_law_1.01,0.1525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,power_law_1.01,0.20614399909973144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,power_law_1.01,0.2568896055221558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,power_law_1.01,0.34023680686950686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,power_law_1.01,0.44357762336730955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,power_law_1.01,0.6358335971832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,power_law_1.01,0.7868671894073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,balanced,0.047397335370381675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,balanced,0.07521066566308339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,balanced,0.08156266808509827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,balanced,0.08448533217112224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,balanced,0.08182933429876964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,balanced,0.08409066994984944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,balanced,0.0844693382581075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,balanced,0.08555733164151509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,balanced,0.08570667107899983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,balanced,0.08930133779843648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,balanced,0.08939199646313985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,balanced,0.09311999877293904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,balanced,0.09554133812586467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,balanced,0.09903466701507568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,balanced,0.11135466893513997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,balanced,0.12995200355847678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,balanced,0.13953066865603128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,balanced,0.17694399754206339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,balanced,0.1980266571044922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,balanced,0.26845866441726685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,balanced,0.31455467144648236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,balanced,0.42900268236796063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,balanced,0.5430026849110922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,power_law_1.2,0.023647999763488768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,power_law_1.2,0.023161600530147552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,power_law_1.2,0.022924800217151643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,power_law_1.2,0.022470399737358093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,power_law_1.2,0.023468799889087677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,power_law_1.2,0.024403199553489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,power_law_1.2,0.025779199600219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.0769599974155426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.0820032000541687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.12277120351791382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.1443071961402893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.06981760263442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.08724480271339416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.09584640264511109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.1488960027694702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.18120959997177125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.2040191888809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.2821824073791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.3604543924331665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,power_law_1.2,0.4992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,power_law_1.2,0.7477888107299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.19304959774017333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.1359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.0953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.09214720129966736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.09402880072593689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.11743999719619751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.13344000577926635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.1310912013053894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.15781760215759277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.18519680500030516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.2244800090789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.2525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.3089792013168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,power_law_1.01,0.3961087942123413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,power_law_1.01,0.6141056060791016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,power_law_1.01,0.714188814163208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.158406400680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.20105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.22675840854644774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.23457279205322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.21984000205993653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.27704319953918455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.2938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.2752959966659546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.28273921012878417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.28499839305877683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.2939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.3092736005783081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.2946367979049683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.3257983922958374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.3405695915222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.35420799255371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.3686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.4115776062011719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.4561215877532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.5963071823120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,power_law_1.2,0.6350272178649903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,power_law_1.2,0.7437056064605713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,power_law_1.2,0.7136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,power_law_1.2,0.9482048034667969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,power_law_1.2,1.1103167533874512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,power_law_1.2,1.6016895294189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,power_law_1.2,2.430796813964844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.24828801155090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.08509439826011658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.09040639996528625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.09511680006980897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.11382399797439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.12016639709472657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.1751487970352173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.22736639976501466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.3378943920135498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.41106557846069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,power_law_1.2,0.5993216037750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,power_law_1.2,0.7748799800872803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,power_law_1.01,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,power_law_1.01,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,power_law_1.01,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,power_law_1.01,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,power_law_1.01,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,power_law_1.01,0.11196160316467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,power_law_1.01,0.11657600402832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,power_law_1.01,0.12817280292510985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,power_law_1.01,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,power_law_1.01,0.18465919494628907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,power_law_1.01,0.2420736074447632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,power_law_1.01,0.29282560348510744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,power_law_1.01,0.40176000595092776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,power_law_1.01,0.5011136054992675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,power_law_1.01,0.7083968162536621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,power_law_1.01,0.9218879699707031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,power_law_1.01,1.338003158569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,power_law_1.01,1.7525247573852538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,balanced,0.035743998984495796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,balanced,0.0379573330283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,balanced,0.040234667559464775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,balanced,0.07017600039641063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,balanced,0.09269332885742188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,balanced,0.10102933645248413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,balanced,0.1274666686852773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,balanced,0.14706666270891824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,balanced,0.19889599084854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,balanced,0.24873065948486328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,balanced,0.34539198875427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,balanced,0.4369279940923055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.2,0.08877440094947815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.2,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.2,0.08451200127601624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.2,0.09073920249938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.2,0.08651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.2,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.2,0.09045760035514831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.2,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.2,0.09544320106506347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.2,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.2,0.09752320051193238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.2,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.2,0.10473599433898925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.2,0.11317119598388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.2,0.13190399408340453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.2,0.14346879720687866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.2,0.1801344037055969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.2,0.21305599212646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.2,0.27075839042663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.2,0.3268480062484741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.2,0.43111681938171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.2,0.5261184215545655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.2,0.7416959762573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.09031680226325989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.10866559743881225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.1302847981452942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.16833920478820802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.20428159236907958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.26389760971069337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.3243583917617798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.464140796661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,power_law_1.01,0.6122367858886719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.2,0.9069631576538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,balanced,0.03704000016053518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,balanced,0.034858666360378265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,balanced,0.05656533439954122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,balanced,0.0798880010843277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,balanced,0.08813866972923279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,balanced,0.10322667161623637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,balanced,0.125791996717453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,balanced,0.07075733443101247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,balanced,0.09609599908192952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,balanced,0.12838400403658548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,balanced,0.1973759929339091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,balanced,0.3321653405825297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,balanced,0.40028266112009686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,balanced,0.4043733278910319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,balanced,0.40802133083343506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,balanced,0.4114346504211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,balanced,0.41352001825968426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,balanced,0.41523198286692303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,balanced,0.4201493263244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,balanced,0.4248746633529663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,balanced,0.431717316309611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,balanced,0.4417813221613566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,balanced,0.4498506784439087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,balanced,0.469215989112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,balanced,0.5076800187428793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.5478506485621134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.8132533232371012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.6942666371663412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.9890720049540201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,balanced,1.0427573521931965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,balanced,1.5622612635294597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,balanced,1.8130720456441243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,balanced,2.611973285675049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,balanced,3.3574241002400718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,power_law_1.01,0.029017600417137145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,power_law_1.01,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,power_law_1.01,0.02887679934501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,power_law_1.01,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,power_law_1.01,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,power_law_1.01,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,power_law_1.01,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.01,0.12919679880142212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.01,0.15038080215454103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.01,0.2132863998413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.01,0.2512128114700317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.01,0.35494399070739746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.01,0.5954432010650634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.01,0.9141823768615722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.2,0.12443519830703735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.2,0.15059839487075805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.2,0.18732160329818726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.2,0.25169920921325684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.2,0.31182079315185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.2,0.46156158447265627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.2,0.501964807510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.2,0.5097023963928222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.2,0.52641921043396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.2,0.5346496105194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.2,0.5545407772064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.01,1.3071616172790528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.2,0.5663871765136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.2,0.5861631870269776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.2,0.6061312198638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.2,0.6324480056762696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.2,0.6414527893066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.2,0.683622407913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.2,0.7599167823791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.2,0.8304703712463379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.2,0.9969535827636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.2,1.1098048210144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.2,1.3648256301879882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.2,1.6574272155761718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.2,2.233132743835449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.2,2.819820785522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.2,3.9364990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.2,4.899059295654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.2,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.2,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.2,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.2,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.2,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.2,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.2,0.10446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.2,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.2,0.14860160350799562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.2,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.2,0.24870400428771972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.2,0.3671488046646118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.2,0.43482241630554197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,power_law_1.01,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,power_law_1.01,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,power_law_1.01,0.06980479955673217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,power_law_1.01,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,power_law_1.01,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,power_law_1.01,0.08593279719352723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,power_law_1.01,0.09686400294303894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,power_law_1.01,0.10974080562591552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.01,0.1276352047920227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.01,0.15313279628753662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.01,0.16830079555511473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.01,0.2132352113723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.01,0.24531838893890381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.01,0.3475775957107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.01,0.43751678466796873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.01,0.6479487895965577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.01,0.9070976257324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.11902079582214356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.1552191972732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.1803712010383606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.24703359603881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.3052031993865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.4319744110107422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.5637119770050049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.04190079867839813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,power_law_1.2,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,power_law_1.2,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,power_law_1.2,0.04449920058250427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,power_law_1.2,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,power_law_1.2,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,power_law_1.2,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,power_law_1.2,0.07159680128097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,power_law_1.2,0.07487360239028931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,power_law_1.2,0.08382080197334289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,power_law_1.2,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.09504640102386475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.2,0.10121599435806275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.2,0.12014080286026001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.2,0.1423807978630066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.2,0.1801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.10198400020599366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.2,0.20994560718536376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.2,0.3292351961135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.2,0.4202303886413574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.13177599906921386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.15943679809570313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.20841600894927978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.26591360569000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.2,0.5878528118133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,power_law_1.2,0.3342655897140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,power_law_1.2,0.421292781829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.2,0.8501184463500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,power_law_1.2,0.7487552165985107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,balanced,0.06037333110968272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,balanced,0.06857599814732869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,balanced,0.0806933343410492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,balanced,0.10533866286277771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,balanced,0.1477120021979014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,balanced,0.15053332845369974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,balanced,0.14863466223080954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,balanced,0.14852266510327658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,balanced,0.15272000432014465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,balanced,0.15533333023389181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,balanced,0.15943466623624167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,balanced,0.15858667095502219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,balanced,0.16270400087038675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,balanced,0.17059199015299478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,balanced,0.17266666889190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,balanced,0.18285866578420004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,balanced,0.20071999231974283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,balanced,0.21791466077168783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,balanced,0.2581706643104553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,balanced,0.29367466767628986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,balanced,0.38072534402211505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,balanced,0.4648053248723348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,balanced,0.6758560339609782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,balanced,0.8617013295491537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,balanced,1.2546772956848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,balanced,1.6552426020304363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,power_law_1.2,0.7299967765808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,balanced,0.0320266659061114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,balanced,0.034858666360378265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,balanced,0.05472533404827118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,balanced,0.08100800216197968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,balanced,0.12231466174125671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,balanced,0.12576533357302347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,balanced,0.12738666931788126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,balanced,0.12818666299184164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,balanced,0.12981866796811423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,balanced,0.13158933321634927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,balanced,0.13010666767756143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,balanced,0.13638400038083395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,balanced,0.1352000037829081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,balanced,0.14012266198794046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,balanced,0.14999999602635702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,balanced,0.15178133050600687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,balanced,0.1585546632607778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,balanced,0.17420266071955362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,balanced,0.19002666076024374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,balanced,0.23057599862416586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,balanced,0.2537279923756917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,balanced,0.3293013374010722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,balanced,0.39347732067108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,balanced,0.5511519908905029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,balanced,0.6865493456522623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,balanced,0.9732159773508707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,balanced,1.266213337580363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,balanced,0.031744000812371574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,balanced,0.03323733309904734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,balanced,0.04048000027736028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,balanced,0.04517866671085358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,balanced,0.10321600238482158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,balanced,0.13743467132250467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,balanced,0.16742400328318277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,balanced,0.06121066709359487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,balanced,0.07933866480986278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,balanced,0.09406933188438416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,balanced,0.10172800223032634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,balanced,0.10329066713651021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,balanced,0.10533333818117778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,balanced,0.10566932956377666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,balanced,0.10588799913724263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,balanced,0.10610666871070862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,balanced,0.10763733585675557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,balanced,0.109333336353302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,balanced,0.11557333668073018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,balanced,0.1183519959449768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,balanced,0.12142933408419292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,balanced,0.1270133356253306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,balanced,0.13398399949073792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,balanced,0.15010133385658264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,balanced,0.16505066553751627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,balanced,0.2180053393046061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,balanced,0.25299733877182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,balanced,0.3530293305714925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,balanced,0.4556106726328532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,balanced,0.6150826613108317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,balanced,0.8152799606323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,balanced,0.050341332952181496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,balanced,0.05583466589450836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,balanced,0.06633066634337108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,balanced,0.0786293347676595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,balanced,0.0886400043964386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,balanced,0.10896533727645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,balanced,0.12826133767763773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,balanced,0.1897546648979187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,balanced,0.23964800437291464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,balanced,0.3349226713180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,balanced,0.42797334988911945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,balanced,0.030847998956839245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,balanced,0.03019733230272929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,balanced,0.03383466601371765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,balanced,0.03663466622432073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,balanced,0.1032480001449585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,balanced,0.14199999968210855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,balanced,0.1749066710472107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,balanced,0.24713599681854248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,balanced,0.3094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,balanced,0.09586133559544881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,balanced,0.14645333091417947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,balanced,0.24567999442418417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,balanced,0.24503467480341592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,balanced,0.2466986576716105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,balanced,0.24668800830841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,balanced,0.25033066670099896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,balanced,0.2504799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,balanced,0.25252799193064374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,balanced,0.2532106637954712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,balanced,0.2553760011990865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,balanced,0.2589120070139567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,balanced,0.26124266783396405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,balanced,0.273909330368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,balanced,0.2877279917399089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,balanced,0.2921706636746724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,balanced,0.30321600039800006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,balanced,0.3744586706161499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.3914666573206584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.5060266653696696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.5871306657791138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,balanced,0.8023253281911215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,balanced,1.0069279670715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,balanced,1.463088035583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,balanced,1.8749173482259114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,balanced,2.7240158716837564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,balanced,3.5950133005777993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.10988160371780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.15306240320205688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.18911999464035034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.31432960033416746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.32408320903778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.3455744028091431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.35869441032409666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.36036479473114014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.3791232109069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.374726390838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.38757760524749757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.40947837829589845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.42239999771118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.01,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.42456960678100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.01,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.01,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.01,0.118777596950531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.01,0.172435200214386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.01,0.1869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.01,0.19089280366897582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.01,0.19503999948501588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.46981120109558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.01,0.19630080461502075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.48404479026794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.01,0.2031872034072876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.5577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.01,0.20954880714416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.6105408191680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.01,0.2065216064453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.01,0.2207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.01,0.23220479488372803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.01,0.24523520469665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.01,0.24048640727996826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.01,0.26549758911132815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,power_law_1.2,0.7411647796630859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.01,0.26973440647125246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.01,0.3053056001663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,power_law_1.2,0.8783359527587891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.01,0.3392512083053589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.01,0.40574078559875487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.01,0.44399361610412597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.01,0.5708479881286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,power_law_1.2,1.1566207885742188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.01,0.7028031826019288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.01,0.9570240020751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.01,1.183232021331787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,power_law_1.2,1.4126591682434082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,balanced,0.0688213308652242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,balanced,0.09598400195439656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,balanced,0.14570666352907816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,balanced,0.2433919906616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,balanced,0.2446613311767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,balanced,0.24498132864634195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,balanced,0.2469600041707357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,balanced,0.2471146583557129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,balanced,0.24567999442418417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,balanced,0.247706671555837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,balanced,0.2493013342221578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,balanced,0.24861333767573038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,balanced,0.2560853362083435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,balanced,0.25569599866867065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,balanced,0.2592853307723999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,balanced,0.26658666133880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,balanced,0.2723466753959656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,balanced,0.28355199098587036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,balanced,0.3089866638183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,balanced,0.33393601576487225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,balanced,0.35711999734242755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,balanced,0.4377760092417399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,balanced,0.47171199321746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,balanced,0.6207733154296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,balanced,0.6990773677825928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,power_law_1.2,1.9346431732177733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,power_law_1.2,2.447756767272949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,power_law_1.2,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,power_law_1.2,0.10840959548950195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,power_law_1.2,0.16425600051879882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,power_law_1.2,0.2185983896255493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,power_law_1.2,0.2808063983917236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,power_law_1.2,0.4088575839996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,power_law_1.2,0.43595519065856936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,power_law_1.2,0.4481088161468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,power_law_1.2,0.45469441413879397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,power_law_1.2,0.4780928134918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,power_law_1.2,0.4963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,power_law_1.2,0.5095615863800049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,power_law_1.2,0.5268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,power_law_1.2,0.5511807918548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,power_law_1.2,0.5695104122161865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,power_law_1.2,0.6001215934753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,power_law_1.2,0.672217607498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,power_law_1.2,0.6777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,power_law_1.2,0.7809408187866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,power_law_1.2,0.8354304313659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,power_law_1.2,1.0003007888793944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,power_law_1.2,1.1077823638916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,power_law_1.2,1.4232959747314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,power_law_1.2,1.7100671768188476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,power_law_1.2,2.3073535919189454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,power_law_1.2,2.9153984069824217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,balanced,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,balanced,0.07918933530648549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,balanced,0.11673600474993388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,balanced,0.17895466089248657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,balanced,0.18172266085942587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,balanced,0.18241065740585327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,balanced,0.18530132373174033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,balanced,0.18401066462198892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,balanced,0.18996800978978476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,balanced,0.1914773384730021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,balanced,0.1904266675313314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,balanced,0.19754666090011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,balanced,0.20404799779256186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,balanced,0.20613867044448853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,balanced,0.21601066986719766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,balanced,0.23291732867558798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,balanced,0.24453866481781006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,balanced,0.2771413326263428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,balanced,0.30927467346191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,balanced,0.39021865526835126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,balanced,0.45001598199208576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,balanced,0.6020693381627401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,balanced,0.727674643198649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,balanced,1.0357226530710857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,balanced,1.3014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,power_law_1.2,0.04472959935665131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,power_law_1.2,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,power_law_1.2,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,power_law_1.2,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,power_law_1.2,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,power_law_1.2,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,power_law_1.2,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,power_law_1.2,0.07493119835853576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,power_law_1.2,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,power_law_1.2,0.09196159839630128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,power_law_1.2,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,power_law_1.2,0.10977920293807983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,power_law_1.2,0.12974079847335815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,power_law_1.2,0.1420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,power_law_1.2,0.17260160446166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,power_law_1.2,0.20315520763397216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,power_law_1.2,0.26421759128570554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,power_law_1.2,0.3344063997268677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,power_law_1.2,0.470304012298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,power_law_1.2,0.5885632038116455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,power_law_1.2,0.8656064033508301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,power_law_1.2,1.1443327903747558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,power_law_1.01,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,power_law_1.01,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,power_law_1.01,0.046239998936653134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,power_law_1.01,0.05372160077095032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,power_law_1.01,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,power_law_1.01,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,power_law_1.01,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.01,0.09571200013160705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.01,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.01,0.14792319536209106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.01,0.17779840230941774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.01,0.24277119636535643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.01,0.2930880069732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.01,0.4525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.01,0.5988160133361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,power_law_1.2,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,power_law_1.2,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,power_law_1.2,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,power_law_1.2,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,power_law_1.2,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,power_law_1.2,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,power_law_1.2,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,power_law_1.2,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,power_law_1.2,0.063673597574234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,power_law_1.2,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,power_law_1.2,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,power_law_1.2,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,power_law_1.2,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,power_law_1.2,0.15811840295791627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,power_law_1.2,0.21057279109954835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,power_law_1.2,0.2671744108200073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,power_law_1.2,0.3824448108673096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,power_law_1.2,0.5022528171539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,power_law_1.2,0.7027711868286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,power_law_1.2,0.9310976028442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.17976959943771362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.2027008056640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.28342399597167967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.3930752038955688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.5977727890014648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.6712128162384033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.12706559896469116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.13873280286788942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.16934399604797362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.2006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.2412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.34828801155090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.3842047929763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.39066879749298095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.40988798141479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.4110720157623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.4156032085418701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.43559041023254397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.4443967819213867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.45912961959838866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.4696959972381592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.47134718894958494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.49767041206359863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.5713024139404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.6187392234802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.7488639831542969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.8604864120483399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,1.1247679710388183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.368006420135498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.922105598449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,2.413516807556152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,3.5020416259765623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,4.584998321533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,power_law_1.01,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,power_law_1.01,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,power_law_1.01,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,power_law_1.01,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,power_law_1.01,0.05191680192947388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,power_law_1.01,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.01,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.01,0.10970879793167114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.01,0.12783360481262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.01,0.16963839530944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.01,0.20848000049591064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.01,0.31781759262084963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.01,0.4354368209838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.01,0.6802112102508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.01,0.7965248107910157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,balanced,0.08750399947166443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,balanced,0.12560000022252402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,balanced,0.20048532883326212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,balanced,0.3370506763458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,balanced,0.46501866976420086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,balanced,0.4702986478805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,balanced,0.47229333718617755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,balanced,0.47413865725199383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,balanced,0.47572267055511475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,balanced,0.48137064774831134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,balanced,0.4880106846491496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,balanced,0.4917440017064412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,balanced,0.49929598967234295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,balanced,0.509498675664266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,balanced,0.5184266567230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,balanced,0.5365493297576904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,balanced,0.6193386713663737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,balanced,0.6180853446324667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,balanced,0.8683199882507324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,balanced,0.8528052965799967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,balanced,1.332634687423706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,balanced,1.3666720390319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,balanced,2.053925355275472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,balanced,2.309941291809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,balanced,3.462090810139974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,balanced,4.2628374099731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,power_law_1.01,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,power_law_1.01,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,power_law_1.01,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,power_law_1.01,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,power_law_1.01,0.10392960309982299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,power_law_1.01,0.10350719690322877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,power_law_1.01,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,power_law_1.01,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,power_law_1.01,0.11174399852752685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,power_law_1.01,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,power_law_1.01,0.11668479442596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,power_law_1.01,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,power_law_1.01,0.1194815993309021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,power_law_1.01,0.12463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,power_law_1.01,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,power_law_1.01,0.13345279693603515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,power_law_1.01,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,power_law_1.01,0.1650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,power_law_1.01,0.17727999687194823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,power_law_1.01,0.21683199405670167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,power_law_1.01,0.22643840312957764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,power_law_1.01,0.2970304012298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,power_law_1.01,0.3556864023208618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,power_law_1.01,0.4832128047943115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,power_law_1.01,0.6015679836273193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,power_law_1.01,0.8666687965393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,power_law_1.01,1.0867456436157226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.2,0.09973120093345642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.2,0.09613440036773682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.2,0.0940671980381012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.2,0.10081280469894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.2,0.09958400130271912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.2,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.2,0.09522560238838196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.2,0.09978880286216736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.2,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.2,0.09979519844055176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.2,0.104038405418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.2,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.2,0.10717439651489258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.2,0.11105279922485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.2,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.2,0.1215872049331665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.2,0.12441600561141967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.2,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.2,0.15774719715118407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.2,0.2076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.2,0.2309056043624878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.2,0.29706881046295164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.2,0.2860415935516357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.2,0.40945920944213865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.2,0.5022079944610596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.2,0.7026879787445068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.2,0.952518367767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,balanced,0.034058667719364166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,balanced,0.038533332447210945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,balanced,0.04491200049718221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,balanced,0.06777599950631459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,balanced,0.0685280015071233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,balanced,0.09239466985066731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,balanced,0.12365333239237468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,balanced,0.13620799779891968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,balanced,0.19020267327626547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,balanced,0.21331733465194702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,balanced,0.0674720009167989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,balanced,0.14193600416183472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,balanced,0.24515734116236368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,balanced,0.4368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,balanced,0.8218932946523031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,balanced,0.8225440184275309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,balanced,0.8289013703664144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,balanced,0.8292746543884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,balanced,0.8329973220825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,balanced,0.8331147034962972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,balanced,0.8386987050374349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,balanced,0.8448266983032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,balanced,0.8402720292409261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,balanced,0.8545653025309244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,balanced,0.854090690612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,balanced,0.8608373006184896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,balanced,0.879957358042399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,balanced,0.8957440058390299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,balanced,0.9249653021494547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,balanced,0.9676853020985922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,balanced,1.0172639687856038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,balanced,1.0637013117472331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,balanced,1.3479785919189453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,balanced,1.3442079226175945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,balanced,1.8007200558980305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,balanced,2.0011466344197593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,balanced,0.03802666564782461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,balanced,0.04032533367474874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,balanced,0.05433600147565206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,balanced,0.05625600119431814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,balanced,0.05643733342488607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,balanced,0.05806933343410492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,balanced,0.08085866769154866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,balanced,0.10429333647092183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,balanced,0.11359999577204387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,balanced,0.15223999818166098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,balanced,0.17002665996551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,balanced,0.23403199513753256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,balanced,0.2770026723543803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,balanced,0.3684320052464803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,balanced,0.4724160035451253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,power_law_1.01,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,power_law_1.01,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,power_law_1.01,0.11125760078430176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,power_law_1.01,0.14420479536056519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,power_law_1.01,0.17306239604949952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,power_law_1.01,0.2263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,power_law_1.01,0.23215360641479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,power_law_1.01,0.24513280391693115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,power_law_1.01,0.2488192081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,power_law_1.01,0.2573951959609985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,power_law_1.01,0.2728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,power_law_1.01,0.28089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,power_law_1.01,0.29804799556732176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,power_law_1.01,0.3180351972579956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,power_law_1.01,0.3331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,power_law_1.01,0.3439872026443481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,power_law_1.01,0.39857280254364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,power_law_1.01,0.44573440551757815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,power_law_1.01,0.5548736095428467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,power_law_1.01,0.6164224147796631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,power_law_1.01,0.79137282371521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,power_law_1.01,0.9726847648620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,power_law_1.01,1.3075648307800294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,power_law_1.01,1.6408000946044923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,power_law_1.01,2.3098175048828127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,power_law_1.01,2.9944576263427733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,power_law_1.01,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,power_law_1.01,0.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,power_law_1.01,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,power_law_1.01,0.07936639785766601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,power_law_1.01,0.07902079820632935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,power_law_1.01,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,power_law_1.01,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,power_law_1.01,0.10188159942626954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,power_law_1.01,0.1032256007194519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,power_law_1.01,0.1068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,power_law_1.01,0.11674879789352417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,power_law_1.01,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,power_law_1.01,0.1562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,power_law_1.01,0.18135679960250856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,power_law_1.01,0.2542975902557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,power_law_1.01,0.2973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,power_law_1.01,0.3775487899780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,power_law_1.01,0.5003583908081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,power_law_1.01,0.6779007911682129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,power_law_1.01,0.9736063957214356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,power_law_1.01,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,power_law_1.01,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,power_law_1.01,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,power_law_1.01,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,power_law_1.01,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,power_law_1.01,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,power_law_1.01,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,power_law_1.01,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,power_law_1.01,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,power_law_1.01,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.01,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.01,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.01,0.13389439582824708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.01,0.15936000347137452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.01,0.21976959705352783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.01,0.26915841102600097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.01,0.37652480602264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.01,0.46720638275146487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.2052288055419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.32512640953063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.3376960039138794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.3459199905395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.3521536111831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.3584511995315552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.357094407081604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.35167360305786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.3696768045425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.3914367914199829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.41096959114074705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.4192959785461426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.4224063873291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.4702655792236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.477459192276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.5349120140075684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,power_law_1.2,0.5800511837005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,power_law_1.2,0.7068352222442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,power_law_1.2,0.7883456230163575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,power_law_1.2,1.030348777770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,power_law_1.2,1.251699161529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,power_law_1.2,1.653811264038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,balanced,0.07235733171304067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,balanced,0.09117866555849712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,balanced,0.10774933298428853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,balanced,0.1106666624546051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,balanced,0.10940266648928325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,balanced,0.10654399792353313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,balanced,0.10679999987284343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,balanced,0.10717866818110149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,balanced,0.10923199852307637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,balanced,0.10918399691581726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,balanced,0.11089600125948589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,balanced,0.11594667037328084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,balanced,0.1156213382879893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,balanced,0.11779200037320454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,balanced,0.14669332901636759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,balanced,0.13587199648221335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,balanced,0.15946666399637857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,balanced,0.16962667306264242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,balanced,0.22998400529225668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,balanced,0.20525866746902466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,balanced,0.26342399915059406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,balanced,0.280949334303538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,balanced,0.383568008740743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,balanced,0.4301760196685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,power_law_1.2,2.0223295211791994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,power_law_1.2,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,power_law_1.2,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,power_law_1.2,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,power_law_1.2,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,power_law_1.2,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,power_law_1.2,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,power_law_1.2,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,power_law_1.2,0.10808320045471191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.2,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.2,0.13194880485534669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.2,0.1590208053588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.2,0.2207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.2,0.27907838821411135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.2,0.38172159194946287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.2,0.5186816215515136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.2,0.7880256175994873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.2,1.08603515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.0967296004295349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.17467520236968995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.17660800218582154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.20388479232788087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.1980031967163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,power_law_1.2,0.0787392020225525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.22172160148620607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.20906879901885986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.20633599758148194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,power_law_1.2,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.23288960456848146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.2211519956588745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,power_law_1.2,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.22432639598846435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.21790080070495604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.23235199451446534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.23618559837341307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,power_law_1.2,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.2435391902923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,power_law_1.2,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.2672575950622559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,power_law_1.2,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,power_law_1.2,0.06362879872322083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,power_law_1.2,0.07527040243148804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,power_law_1.2,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.294265604019165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.28587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,power_law_1.2,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.32610559463500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,power_law_1.2,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.3451263904571533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,power_law_1.2,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.4176896095275879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,power_law_1.2,0.15638400316238404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,power_law_1.01,0.47969918251037597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,power_law_1.2,0.2066495895385742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,power_law_1.2,0.28193280696868894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,power_law_1.01,0.601907205581665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,power_law_1.2,0.3917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,power_law_1.01,0.6813504219055175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,power_law_1.2,0.5400832176208497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,power_law_1.01,0.913548755645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,power_law_1.01,1.099283218383789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.14545919895172119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.19435520172119142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.1912511944770813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.19409919977188111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.202675199508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.19833600521087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.21367039680480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.2165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.21943678855895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.23168001174926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.23854079246520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.252889609336853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.2792896032333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.30150399208068845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.3802175998687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.39473280906677244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.4893375873565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.587718391418457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,0.7527488231658935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.0222528457641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,1.3162688255310058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,1.8530048370361327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,power_law_1.01,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,power_law_1.01,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,power_law_1.01,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,power_law_1.01,0.07273600101470948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,power_law_1.01,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,power_law_1.01,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,power_law_1.01,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,power_law_1.01,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,power_law_1.01,0.08483200073242188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,power_law_1.01,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,power_law_1.01,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,power_law_1.01,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,power_law_1.01,0.10046080350875855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,power_law_1.01,0.10515199899673462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,power_law_1.01,0.115392005443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.01,0.12288000583648681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.01,0.13999999761581422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.01,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.01,0.19857920408248902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.01,0.25155839920043943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.01,0.3439167976379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.01,0.4023871898651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.01,0.6108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.01,0.771827220916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,balanced,0.04849066833655039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,balanced,0.056661332647005715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,balanced,0.0621973325808843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,balanced,0.06803733110427856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,balanced,0.0754559983809789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,balanced,0.07669333120187123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,balanced,0.09578667084376018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,balanced,0.10094400246938069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,balanced,0.12743999560674033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,balanced,0.14189866185188293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,balanced,0.18372267484664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,balanced,0.1794346570968628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,balanced,0.22991466522216797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,balanced,0.27082133293151855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,balanced,0.36215468247731525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,balanced,0.4440000057220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,power_law_1.01,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,power_law_1.01,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,power_law_1.01,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,power_law_1.01,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,power_law_1.01,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,power_law_1.01,0.1020095944404602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,power_law_1.01,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,power_law_1.01,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,power_law_1.01,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,power_law_1.01,0.11040639877319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,power_law_1.01,0.11893119812011718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,power_law_1.01,0.12239359617233277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,power_law_1.01,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,power_law_1.01,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.01,0.14696320295333862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.01,0.17422720193862914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.01,0.20456960201263427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.01,0.27638399600982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.01,0.3532288074493408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.01,0.46749439239501955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.01,0.6060927867889404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.01,0.8540224075317383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.01,1.1576576232910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,balanced,0.07643733421961467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,balanced,0.10042132933934529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,balanced,0.10507733623186748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,balanced,0.1025279959042867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,balanced,0.10166399677594502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,balanced,0.10061333576838176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,balanced,0.10001066327095032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,balanced,0.10099732875823975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,balanced,0.11269866426785786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,balanced,0.10857066512107849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,balanced,0.11417599519093831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,balanced,0.1197653313477834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.12598400314648947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.14401599764823914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.1575040022532145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.1883253256479899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.21709332863489786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.2763413389523824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.3468853235244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.46253331502278644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.5929706494013468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,power_law_1.2,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,power_law_1.2,0.10730880498886108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,power_law_1.2,0.13676799535751344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,power_law_1.2,0.16769920587539672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,power_law_1.2,0.23176319599151612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,power_law_1.2,0.24606719017028808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,power_law_1.2,0.25920639038085935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,power_law_1.2,0.2613696098327637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,power_law_1.2,0.276364803314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,power_law_1.2,0.2856064081192017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,power_law_1.2,0.29395840167999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,power_law_1.2,0.30741119384765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,power_law_1.2,0.3164031982421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,power_law_1.2,0.3261823892593384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,power_law_1.2,0.3467200040817261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,power_law_1.2,0.3958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,power_law_1.2,0.41393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,power_law_1.2,0.4955264091491699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,power_law_1.2,0.5444992065429688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,power_law_1.2,0.6939136028289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,power_law_1.2,0.8149824142456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,power_law_1.2,1.0852031707763672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,power_law_1.2,1.3295488357543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,power_law_1.2,1.8433536529541015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,power_law_1.2,2.3433088302612304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,power_law_1.2,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,power_law_1.2,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,power_law_1.2,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,power_law_1.2,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,power_law_1.2,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.2,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.2,0.12166399955749511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.2,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.2,0.21231999397277831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.2,0.2504447937011719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.2,0.4090688228607178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.2,0.456774377822876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.11749119758605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.17123839855194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.2189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.3480448007583618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.3747136116027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.37655680179595946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.38823039531707765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.38687360286712646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.3872319936752319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.3903104066848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.3932032108306885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.3996608018875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.409497594833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.4207744121551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.45522561073303225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.4604288101196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.511359977722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,power_law_1.2,0.5671679973602295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,power_law_1.2,0.6204800128936767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,power_law_1.2,0.7897984027862549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,power_law_1.2,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,power_law_1.2,0.11054719686508178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,power_law_1.2,0.11029759645462037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,power_law_1.2,0.09404159784317016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,power_law_1.2,0.10819200277328492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,power_law_1.2,0.11403520107269287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,power_law_1.2,0.11289600133895875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,power_law_1.2,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,power_law_1.2,0.11891839504241944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,power_law_1.2,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,power_law_1.2,0.8922176361083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,power_law_1.2,0.12670719623565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,power_law_1.2,0.14168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,power_law_1.2,0.14767999649047853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,power_law_1.2,0.15555839538574218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,power_law_1.2,0.2082751989364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,power_law_1.2,0.20283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,power_law_1.2,0.24795520305633545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,power_law_1.2,0.25553278923034667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,power_law_1.2,0.34686079025268557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,power_law_1.2,0.399782395362854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,power_law_1.2,1.1597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,power_law_1.2,0.5126976013183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,power_law_1.2,0.6889920234680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,power_law_1.2,1.464358425140381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,power_law_1.2,0.9620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.2,0.8483455657958985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,power_law_1.2,2.040070343017578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,power_law_1.2,1.178816032409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,power_law_1.2,2.665119934082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,power_law_1.2,3.912070465087891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.2,1.1044672012329102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,power_law_1.2,5.097420883178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.0711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07246080040931702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.11562880277633666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.14202239513397216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.1845695972442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.2187648057937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.3284096002578735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.4016704082489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.5894336223602294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,balanced,0.05439466734727224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,balanced,0.08583466211954753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,balanced,0.12784000237782797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,balanced,0.19809067249298096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,balanced,0.25948800643285114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,balanced,0.26184000571568805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,balanced,0.26497600475947064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,balanced,0.26340800523757935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,balanced,0.2677119970321655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,balanced,0.26994667450586957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,balanced,0.2734453280766805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,balanced,0.27720532814661664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,balanced,0.2845013340314229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,balanced,0.29340267181396484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,balanced,0.30054932832717896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,balanced,0.3129120071729024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,balanced,0.3561226526896159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,balanced,0.3696533441543579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,balanced,0.4968373378117879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,balanced,0.5108213424682617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,balanced,0.7824532985687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,balanced,0.8007520039876302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,balanced,1.1972853342692058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,balanced,1.3213120301564534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,balanced,1.9880480766296387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,balanced,2.4862027168273926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.7462016105651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.1125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.14048000574111938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.1888576030731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.21994879245758056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.24296960830688477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.29270401000976565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.2992768049240112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.3123456001281738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.32814080715179444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.3376895904541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.3476288080215454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.3645375967025757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.37610878944396975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.3901504039764404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.4393343925476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.4791999816894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,1.123417568206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,0.5756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,0.6616447925567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,0.848902416229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,1.0325056076049806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,1.4032383918762208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.01,1.7715839385986327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.01,2.5153024673461912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.01,3.2500415802001954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.4999103546142578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,balanced,0.03443733354409536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.05835199852784475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.07228800157705943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.10770133137702942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.12954666217168173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.1706719994544983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.21751999855041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.05093119740486145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.06464639902114869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.08234879970550538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.08676480054855347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.09927679896354676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.13891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.17870080471038818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.21907200813293456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.3020031929016113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,balanced,0.03469333300987879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,balanced,0.03846399982770284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,balanced,0.03932266682386398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,balanced,0.0383840004603068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,balanced,0.039093332986036934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,balanced,0.05434666574001312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,balanced,0.07063999772071838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,balanced,0.09307199716567993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,balanced,0.12898133198420206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,balanced,0.1546293298403422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,balanced,0.20788266261418661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,balanced,0.26333866516749066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.3841279983520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.5396992206573487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.6794559955596924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,1.038252830505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.3572863578796386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,balanced,0.038431999584039055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,balanced,0.03859733293453852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,balanced,0.040752001106739044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,balanced,0.042064001162846885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,balanced,0.060421332716941833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,balanced,0.06213866670926412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,balanced,0.06444266438484192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,balanced,0.06844800213972728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,balanced,0.07062933345635732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,balanced,0.07987200220425923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,balanced,0.08851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,balanced,0.09520533680915833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,balanced,0.10991999506950378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,balanced,0.1239413321018219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,balanced,0.15929599603017172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,balanced,0.1855093240737915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,balanced,0.24274667104085287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,balanced,0.3049866755803426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,power_law_1.01,0.028249600529670717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,power_law_1.01,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,power_law_1.01,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,power_law_1.01,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,power_law_1.01,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.12727680206298828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.1634752035140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.20371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,power_law_1.01,0.27125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,power_law_1.01,0.37763199806213377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.06889600157737732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.08356480002403259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.08614400029182434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.12684799432754518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.15972479581832885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.1739456057548523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.2284032106399536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.2617088079452515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.3982656002044678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.04283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.07288320064544677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,balanced,0.05751466751098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,balanced,0.05770133435726166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,balanced,0.06066133578618368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,balanced,0.061349332332611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,balanced,0.062394668658574425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,balanced,0.06268799801667531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.11333760023117065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.14488960504531861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.18447999954223632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.23900160789489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.2815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.5172927856445313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.4050623893737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,balanced,0.06297599772612254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.5603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,balanced,0.07020799815654755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,balanced,0.07478400071461995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,balanced,0.0788213312625885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.09653866291046143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.10759466886520386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.13702399532000223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.16075199842453003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.2057973345120748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.24968000253041586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.34302401542663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.4304373264312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,0.8038911819458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.0847359657287599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,1.6121791839599608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,power_law_1.2,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,power_law_1.2,0.11328639984130859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,power_law_1.2,0.144268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,power_law_1.2,0.18591359853744507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,power_law_1.2,0.24544000625610352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,power_law_1.2,0.24781439304351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,power_law_1.2,0.24046080112457274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,power_law_1.2,0.2561791896820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,power_law_1.2,0.26711039543151854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,power_law_1.2,0.27242240905761717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,power_law_1.2,0.27411201000213625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,power_law_1.2,0.27575039863586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,power_law_1.2,0.2977407932281494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,power_law_1.2,0.29687039852142333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,power_law_1.2,0.3144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,power_law_1.2,0.3503232002258301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,power_law_1.2,0.349619197845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,power_law_1.2,0.40583038330078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,power_law_1.2,0.4342912197113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,power_law_1.2,0.5345471858978271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,power_law_1.2,0.5681344032287597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,power_law_1.2,0.7195136070251464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,power_law_1.2,0.8580096244812012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,power_law_1.2,1.1086912155151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,power_law_1.2,1.464179229736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,2.187148857116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,balanced,0.04734399914741516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,balanced,0.08710400263468425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,balanced,0.11322133739789327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,balanced,0.11606933673222859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,balanced,0.11853866775830586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,balanced,0.11948266625404358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,balanced,0.12109333276748657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,balanced,0.12594133615493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,balanced,0.12801600495974222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,balanced,0.1309760014216105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,balanced,0.1350826621055603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,balanced,0.14057067036628723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,balanced,0.1481760044892629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,balanced,0.1588533322016398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,balanced,0.18219733238220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,balanced,0.20286933581034342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,balanced,0.2486506700515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,balanced,0.3090133269627889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,balanced,0.3980426788330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,balanced,0.4912266731262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,balanced,0.7272746562957764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,balanced,0.9125066598256429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,balanced,1.3382399876912434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,balanced,1.75545072555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,power_law_1.01,0.051520001888275144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,power_law_1.01,0.06117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,power_law_1.01,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,power_law_1.01,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,power_law_1.01,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,power_law_1.01,0.14520959854125975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,power_law_1.01,0.15205119848251342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,power_law_1.01,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,power_law_1.01,0.15962239503860473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,power_law_1.01,0.16840959787368776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,power_law_1.01,0.17484159469604493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,power_law_1.01,0.18702720403671264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,power_law_1.01,0.2004863977432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,power_law_1.01,0.2006335973739624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,power_law_1.01,0.23059840202331544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,power_law_1.01,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,power_law_1.01,0.3213248014450073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,power_law_1.01,0.37750399112701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,power_law_1.01,0.4815104007720947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,power_law_1.01,0.5946559906005859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,power_law_1.01,0.8303808212280274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,power_law_1.01,1.03953275680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,power_law_1.01,1.467142391204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,power_law_1.01,1.9337663650512695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,balanced,0.04469866553942362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,balanced,0.07137600084145863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,balanced,0.09934399525324504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,balanced,0.1488746702671051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,balanced,0.20061333974202475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,balanced,0.20096532503763834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,balanced,0.20241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,balanced,0.2023573319117228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,balanced,0.20297600825627646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,balanced,0.20542399088541666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,balanced,0.2064639925956726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,balanced,0.20706133047739664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,balanced,0.21538132429122925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,balanced,0.21370667219161987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,balanced,0.22217599550882974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,balanced,0.23479467630386353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,balanced,0.2379680077234904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,balanced,0.25781333446502686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,balanced,0.27560534079869586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,balanced,0.31699200471242267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,balanced,0.35901331901550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,balanced,0.4306453466415405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,balanced,0.5143306652704874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,balanced,0.6673440138498942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,balanced,0.8448106447855631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,balanced,0.04451733330885569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,balanced,0.04493333399295807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,balanced,0.05818133552869161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,balanced,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,balanced,0.0724480003118515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,balanced,0.07518399755160014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,balanced,0.07623466849327087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,balanced,0.08281599978605907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,balanced,0.08943466345469157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,balanced,0.1029813289642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,balanced,0.11412266890207927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,balanced,0.1481013298034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,balanced,0.1646666626135508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,balanced,0.20799465974171957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,balanced,0.25253866116205853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,balanced,0.3386400143305461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,balanced,0.44921600818634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,balanced,0.1302720010280609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,balanced,0.13537599643071493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,balanced,0.13422933220863342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,balanced,0.1360053320725759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,balanced,0.13874133427937826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,balanced,0.139957328637441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,balanced,0.14127467075983682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,balanced,0.14236266414324442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,balanced,0.14366400241851807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,balanced,0.14857066671053568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,balanced,0.14703466494878134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,balanced,0.15731199582417807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,balanced,0.16636799772580466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,balanced,0.16795732577641806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,balanced,0.17307200034459433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,balanced,0.21370667219161987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,balanced,0.2204266587893168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,balanced,0.28594134251276654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,balanced,0.3251039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,balanced,0.43372801939646405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,balanced,0.5406186580657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,balanced,0.7766933441162109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,balanced,1.0058186848958333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,balanced,1.4737440745035808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,balanced,1.94487460454305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,balanced,0.09880533814430237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,balanced,0.1502346694469452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,balanced,0.2485706607500712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,balanced,0.24940266211827597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,balanced,0.25042132536570233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,balanced,0.25218133131663006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,balanced,0.25125332673390705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,balanced,0.2540000081062317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,balanced,0.25949867566426593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,balanced,0.2572480042775472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,balanced,0.2648266752560933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,balanced,0.2711519996325175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,balanced,0.27380265792210895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,balanced,0.28311999638875324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,balanced,0.29948800802230835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,balanced,0.3134133418401082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,balanced,0.34882664680480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,balanced,0.3920373519261678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,balanced,0.45028801759084064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,balanced,0.5213546752929688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,balanced,0.6750079790751139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,balanced,0.7797386646270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,balanced,1.105183998743693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,balanced,1.3652480443318684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,power_law_1.2,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,power_law_1.2,0.04366720020771027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,power_law_1.2,0.04510720074176788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,power_law_1.2,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,power_law_1.2,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,power_law_1.2,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,power_law_1.2,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,power_law_1.2,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,power_law_1.2,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,power_law_1.2,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,power_law_1.2,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,power_law_1.2,0.15018240213394166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,power_law_1.2,0.19648000001907348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,power_law_1.2,0.23306879997253419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,power_law_1.2,0.3575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,power_law_1.2,0.4159103870391846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,power_law_1.01,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,power_law_1.01,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,power_law_1.01,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,power_law_1.01,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,power_law_1.01,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,power_law_1.01,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,power_law_1.01,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,power_law_1.01,0.12329599857330323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,power_law_1.01,0.14412800073623658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,power_law_1.01,0.18191360235214232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,power_law_1.01,0.21475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,power_law_1.01,0.29670400619506837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,power_law_1.01,0.375436806678772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,power_law_1.01,0.5511936187744141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,power_law_1.01,0.730944013595581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.1024448037147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.1523327946662903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.2389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.4370175838470459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.6635263919830322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.950432014465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.1451135635375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.5274815559387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.5845631599426269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.6822528839111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.7760639190673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.8313919067382813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,balanced,0.047541335225105286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,balanced,0.06099733213583628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,balanced,0.0767626663049062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,balanced,0.07946666578451793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,balanced,0.07776533563931783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,balanced,0.0790293316046397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,balanced,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,balanced,0.08169066905975342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,balanced,0.08302400012811025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,balanced,0.0888159970442454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,balanced,0.08928533395131429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,balanced,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,balanced,0.0995253324508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,balanced,0.10585066676139832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,power_law_1.2,1.8635583877563477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,balanced,0.12086932857831319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,balanced,0.13784533739089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,balanced,0.18022932608922324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,balanced,0.21707733472188315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,power_law_1.2,1.9363840103149415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,balanced,0.31349867582321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,balanced,0.4049333333969116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,balanced,0.5824053287506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,power_law_1.2,1.9776575088500976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,balanced,0.7623626391092936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.014201545715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.1038080215454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.2927743911743166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.4145151138305665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,power_law_1.2,2.7147968292236326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,power_law_1.2,2.7989248275756835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,power_law_1.2,3.171161651611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,power_law_1.2,3.7351104736328127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,power_law_1.2,4.61841926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,power_law_1.2,4.465036773681641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,power_law_1.2,5.549440002441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,power_law_1.2,7.239417266845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,balanced,0.06715733309586842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,balanced,0.06774933139483134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,balanced,0.0906986693541209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,balanced,0.09624532858530681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,balanced,0.09719467163085938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,balanced,0.0978559950987498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,balanced,0.09872000416119893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,balanced,0.09884799520174663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,balanced,0.09911466638247173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,balanced,0.10405332843462627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,balanced,0.10511466860771179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,balanced,0.10600533088048299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,balanced,0.12428266803423564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,balanced,0.12450666228930156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,balanced,0.1304266651471456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,balanced,0.1497173309326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,balanced,0.1572373310724894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,balanced,0.18856000900268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,balanced,0.21384533246358237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,balanced,0.3064799904823303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,balanced,0.3796693483988444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,balanced,0.5321546792984009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,balanced,0.6930027008056641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,balanced,1.0350933074951172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,balanced,1.3525333404541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.09660800099372864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.12670719623565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.1391808032989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.1561728000640869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.19288959503173828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.19711359739303588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.2550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.3038912057876587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.403110408782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.5489855766296386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.7533567905426025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,0.9848959922790528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,balanced,0.21585599581400552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,balanced,0.3932693401972453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,balanced,0.7445600032806396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,balanced,0.7388853232065836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,balanced,0.730949322382609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,balanced,0.7373759746551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,balanced,0.7401813666025797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,balanced,0.7351093292236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,balanced,0.7393013636271158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,balanced,0.7400266329447428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,balanced,0.7716159820556641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,balanced,0.7502826849619547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,balanced,0.7580693562825521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,balanced,0.8010186354319254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,balanced,0.800981362660726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,balanced,0.8075520197550455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,balanced,0.9199999968210856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,balanced,0.9733546574910482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,balanced,0.9893813133239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,balanced,1.2970399856567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,balanced,1.364496072133382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,balanced,1.8477813402811687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,balanced,2.412933349609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,balanced,3.5323359171549478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,balanced,4.666410764058431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,balanced,6.938165028889974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,balanced,9.241962432861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.07286400198936463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.11413120031356812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.14636160135269166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.16788480281829835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.19813120365142822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.27538559436798093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.3567296028137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.5384128093719482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.7199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,balanced,0.03888533264398575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,balanced,0.039120001097520195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,balanced,0.04020266731580099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,balanced,0.043231998880704246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,balanced,0.04141333450873693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,balanced,0.04214933514595032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.055215999484062195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.09996267159779866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.1155413289864858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.14938666423161825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.18530132373174033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.1170432090759277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.3982144355773927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.05799679756164551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.08776959776878357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.17063039541244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.17751679420471192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.1824447989463806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.18960000276565553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.19464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.205401611328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.21161599159240724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.22470400333404542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.2340224027633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.23861119747161866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.28774399757385255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.34291839599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.38676478862762453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.45128960609436036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.4868800163269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6378047943115235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.7832704067230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.0710399627685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.364031982421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,power_law_1.2,1.9530431747436523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.545497512817383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,balanced,0.05541333556175232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,balanced,0.05787200232346853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,balanced,0.06144533554712931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,balanced,0.06304533282915752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,balanced,0.06568533182144165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,balanced,0.06919999917348225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,balanced,0.0766133318344752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,balanced,0.07692266503969829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,balanced,0.08901333808898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.09269866347312927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.1051573355992635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.1190880040327708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.1406826674938202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.16738667090733847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.21105066935221353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.250709335009257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.34883733590443927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.43187201023101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,power_law_1.2,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,power_law_1.2,0.08778240084648133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,power_law_1.2,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,power_law_1.2,0.11164159774780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,power_law_1.2,0.1239743947982788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,power_law_1.2,0.12583680152893068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,power_law_1.2,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,power_law_1.2,0.12718080282211303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,power_law_1.2,0.1318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,power_law_1.2,0.13096959590911866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,power_law_1.2,0.13396480083465576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,power_law_1.2,0.14080640077590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,power_law_1.2,0.1457856059074402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,power_law_1.2,0.149452805519104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,power_law_1.2,0.1548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,power_law_1.2,0.17169920206069947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.2,0.19812480211257935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.2,0.2474047899246216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.2,0.2864448070526123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.2,0.3975744009017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.2,0.46954879760742185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.2,0.6774784088134765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.2,0.8853055953979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.2,1.283993625640869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.2,1.6291328430175782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.1201856017112732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.15275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.22732160091400147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.2410304069519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.25213439464569093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.25530879497528075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.2606911897659302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.26670079231262206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.26837759017944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.27915520668029786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2881792068481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.30600318908691404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.3163264036178589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.3371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.3903872013092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.4203648090362549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.5001344203948974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.5781184196472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.7424511909484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.8976448059082032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.210745620727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,1.5123968124389648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,balanced,0.0509493350982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,balanced,0.07106133302052815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,balanced,0.09771733482678731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,balanced,0.15205867091814676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,balanced,0.26127467552820843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,balanced,0.31083200375239056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,balanced,0.31483733654022217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,balanced,0.31696534156799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,balanced,0.31753599643707275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,balanced,0.3177493413289388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,balanced,0.3215786616007487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,balanced,0.3245546619097392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,balanced,0.32335466146469116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,balanced,0.33027199904123944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,balanced,0.3370240132013957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,balanced,0.3389279842376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,balanced,0.35051735242207843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,balanced,0.3695146640141805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,balanced,0.3758186499277751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,balanced,0.4602773189544678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,balanced,0.44061867396036786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,balanced,0.6471146742502848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,balanced,0.5699040095011393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,balanced,0.8789920012156168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,balanced,0.8862559795379639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,balanced,1.2371413707733154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,balanced,1.585210641225179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,2.1162879943847654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,2.7160512924194338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.11502079963684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.16488959789276122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.17960959672927856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.18880000114440917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.18915200233459473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.19436800479888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.2093951940536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.21958398818969727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.22539520263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.228985595703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2415424108505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.25289599895477294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.268505597114563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.30833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.3506752014160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.43980798721313474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5040256023406983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.6549439907073975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.8454655647277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1197952270507812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.530790424346924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.0521215438842773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.5331327438354494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,balanced,0.06743466854095459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,balanced,0.09512000282605489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,balanced,0.14660267035166422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,balanced,0.14645866552988687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,balanced,0.14828800161679587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,balanced,0.14994133512179056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,balanced,0.14973333477973938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,balanced,0.1490666667620341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,balanced,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,balanced,0.1504906713962555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,balanced,0.15161599715550741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,balanced,0.15667733550071716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,balanced,0.15826132893562317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,balanced,0.16101333498954773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,balanced,0.16744534174601236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,balanced,0.17418134212493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,balanced,0.18589866161346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,balanced,0.19797333081563315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,balanced,0.2217973272005717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,balanced,0.24316799640655518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,balanced,0.3070613344510396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,balanced,0.3439146677652995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,balanced,0.48283199469248456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,balanced,0.5509599844614664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,power_law_1.01,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,power_law_1.01,0.09722880125045777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,power_law_1.01,0.14767359495162963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,power_law_1.01,0.13068159818649291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,power_law_1.01,0.15808000564575195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,power_law_1.01,0.18424320220947266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,power_law_1.01,0.17079039812088012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,power_law_1.01,0.1642688035964966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,power_law_1.01,0.17256959676742553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,power_law_1.01,0.16146559715270997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,power_law_1.01,0.17708799839019776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,power_law_1.01,0.17732479572296142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,power_law_1.01,0.18803839683532714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,power_law_1.01,0.1906623959541321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,power_law_1.01,0.19415680170059205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,power_law_1.01,0.2048703908920288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,power_law_1.01,0.21937921047210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,power_law_1.01,0.22977919578552247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,power_law_1.01,0.2559936046600342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,power_law_1.01,0.29438080787658694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,power_law_1.01,0.3240384101867676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,power_law_1.01,0.40322561264038087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,power_law_1.01,0.462713623046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,power_law_1.01,0.6172160148620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,power_law_1.01,0.7626368045806885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,power_law_1.01,0.9998656272888183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,power_law_1.01,1.3172351837158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.01,0.14947199821472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.01,0.14732799530029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.01,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.01,0.15294719934463502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.01,0.12714240550994874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.01,0.14584319591522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.01,0.14893440008163453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.01,0.14615039825439452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.01,0.14249600172042848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.01,0.15104000568389891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.01,0.15102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.01,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.01,0.1544000029563904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.01,0.16561919450759888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.01,0.16599680185317994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.01,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.01,0.21217920780181884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.01,0.20227839946746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.01,0.25883519649505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.01,0.3063103914260864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.01,0.37379839420318606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.01,0.3690687894821167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.01,0.4840256214141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.01,0.5683135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.01,0.8761728286743165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.01,1.006105613708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,balanced,0.05977599819501241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,balanced,0.059546664357185364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,balanced,0.061759998401006065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,balanced,0.07186133166154225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,balanced,0.08640000224113464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.09678399562835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.12471466263135274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.16586666305859885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.19347200791041055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.26284799973169964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.34257598718007404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.49062931537628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.6357013384501139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.09233919978141784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.14016000032424927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.13582719564437867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.17583359479904176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.2292543888092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.211244797706604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.22273919582366944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.2146496057510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.19522559642791748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.21870720386505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.2136768102645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.22646400928497315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.23060479164123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.24279680252075195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.24706559181213378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.26880640983581544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.30215039253234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.28993279933929444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.35048959255218504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.37843201160430906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,power_law_1.2,0.42957439422607424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,power_law_1.2,0.4984640121459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,power_law_1.2,0.6669055938720703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,power_law_1.2,0.7503232002258301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,power_law_1.2,1.0486144065856933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,power_law_1.2,1.3049535751342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,power_law_1.01,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,power_law_1.01,0.11626240015029907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,power_law_1.01,0.15429760217666627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,power_law_1.01,0.20509440898895265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,power_law_1.01,0.31137919425964355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,power_law_1.01,0.3370176076889038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,power_law_1.01,0.343667197227478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,power_law_1.01,0.3529279947280884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,power_law_1.01,0.35413761138916017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,power_law_1.01,0.36788480281829833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,power_law_1.01,0.3801215887069702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,power_law_1.01,0.39032959938049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,power_law_1.01,0.3929728031158447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,power_law_1.01,0.4104447841644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,power_law_1.01,0.42872958183288573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,power_law_1.01,0.4375296115875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,power_law_1.01,0.48032641410827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,power_law_1.01,0.5038464069366455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,power_law_1.01,0.5972864151000976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,power_law_1.01,0.6497727870941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,power_law_1.01,0.817471981048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,power_law_1.01,0.9682944297790528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,power_law_1.01,1.2910528182983398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,power_law_1.01,1.6455360412597657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,power_law_1.01,2.2541183471679687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,power_law_1.01,2.9005056381225587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.11092480421066284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.13092479705810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.18680319786071778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.22184319496154786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,power_law_1.01,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,power_law_1.01,0.043705600500106814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,power_law_1.01,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,power_law_1.01,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,power_law_1.01,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,power_law_1.01,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,power_law_1.01,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,power_law_1.01,0.06032639741897583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,power_law_1.01,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,power_law_1.01,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,power_law_1.01,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,power_law_1.01,0.16880639791488647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,power_law_1.01,0.22945919036865234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,power_law_1.01,0.30205440521240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,power_law_1.01,0.47905921936035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,power_law_1.01,0.5850560188293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.3173759937286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.01,0.031455999612808226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.01,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.01,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.01,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.01,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.01,0.12973439693450928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.01,0.15891200304031372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.01,0.20567679405212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.01,0.27248640060424806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,power_law_1.01,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,power_law_1.01,0.1112064003944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,power_law_1.01,0.12520320415496827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,power_law_1.01,0.13844480514526367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,power_law_1.01,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,power_law_1.01,0.1401919960975647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,power_law_1.01,0.15082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,power_law_1.01,0.15488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,power_law_1.01,0.1571776032447815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,power_law_1.01,0.1606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,power_law_1.01,0.16903040409088135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,power_law_1.01,0.17816319465637206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,power_law_1.01,0.18414080142974854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,power_law_1.01,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,power_law_1.01,0.22256639003753662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,power_law_1.01,0.2633280038833618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,power_law_1.01,0.30611839294433596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,power_law_1.01,0.3946304082870483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,power_law_1.01,0.4747776031494141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,power_law_1.01,0.6607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,power_law_1.01,0.8038080215454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,power_law_1.01,1.106771183013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,power_law_1.01,1.453446388244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.40222721099853515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,balanced,0.046122665206591286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,balanced,0.04821333289146423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,balanced,0.04923200110594431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,balanced,0.0613973339398702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,balanced,0.06605866551399231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,balanced,0.06962666908899943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,balanced,0.07256533205509186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,balanced,0.09033067027727763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,balanced,0.1258080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,balanced,0.15718932946523032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,balanced,0.2298240065574646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,balanced,0.29598400990168255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,balanced,0.4360160032908122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,balanced,0.5599786837895712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,balanced,0.8123466968536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,balanced,1.06277330716451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,power_law_1.01,0.028191998600959778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,power_law_1.01,0.027987200021743774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,power_law_1.01,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,power_law_1.01,0.03059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,power_law_1.01,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,power_law_1.01,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,power_law_1.01,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,power_law_1.01,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,power_law_1.01,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.01,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,power_law_1.01,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,power_law_1.01,0.051526397466659546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,power_law_1.01,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,power_law_1.01,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.01,0.1886080026626587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,power_law_1.01,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,power_law_1.01,0.08793600201606751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,power_law_1.01,0.10067199468612671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.01,0.23457279205322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,power_law_1.01,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,power_law_1.01,0.13824000358581542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,power_law_1.01,0.17114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,power_law_1.01,0.1991487979888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,power_law_1.01,0.2921983957290649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.01,0.366483211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,power_law_1.01,0.3835839986801147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.01,0.48105602264404296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,power_law_1.01,0.5164991855621338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.01,0.7497151851654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.01,1.126483154296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,power_law_1.01,0.6785408020019531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.01,0.027315199375152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.01,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.01,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.01,0.02956160008907318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.01,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.01,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.01,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.01,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.01,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.01,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.01,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.01,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.01,0.11205120086669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.01,0.14558720588684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.01,0.14568320512771607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.06373119950294495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.10179840326309204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.12211840152740479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.16194560527801513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.2029695987701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.2547008037567139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.33079679012298585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.5052864074707031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.6944064140319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,balanced,0.07069333394368489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,balanced,0.09966933727264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.051974397897720334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.0868287980556488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.127455997467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,balanced,0.15203733245531717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,balanced,0.25752000013987225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,balanced,0.46533334255218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,balanced,0.5770560105641683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.243833589553833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,balanced,0.5770453214645386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,balanced,0.5795040130615234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,balanced,0.5859359900156657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,balanced,0.5859146515528361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,balanced,0.5877120097478231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,balanced,0.5957760016123453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,balanced,0.5998239914576212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,balanced,0.6063199838002523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,balanced,0.6185760100682577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,balanced,0.6249386469523112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,balanced,0.6431306600570679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,power_law_1.01,0.29034879207611086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,balanced,0.6739733219146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,balanced,0.7050666809082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,balanced,0.8448266983032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,balanced,0.8342933654785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,balanced,1.1716319719950359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,balanced,1.1189813613891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,balanced,1.579605261484782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,balanced,1.827162742614746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,balanced,2.9563945134480796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,balanced,3.44161065419515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.0711679995059967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.0743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.10442240238189697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.11666560173034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.13290879726409913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.1563007950782776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.1925503969192505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.22295041084289552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.30191359519958494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.3769216060638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.5104447841644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.6585343837738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.1403712034225464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.13095680475234986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.13722879886627198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.1392575979232788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.1478656053543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.15054719448089598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.15967999696731566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.1624768018722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.16981760263442994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.17269760370254517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.17815040349960326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.20547199249267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.21203839778900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.24826879501342775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.28480000495910646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.3844095945358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.4496511936187744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,power_law_1.2,0.5410943984985351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,power_law_1.2,0.6869760036468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,power_law_1.2,0.9959424018859864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,power_law_1.2,1.3356351852416992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,power_law_1.2,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,power_law_1.2,0.06373119950294495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,power_law_1.2,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,power_law_1.2,0.11091840267181396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,power_law_1.2,0.1295807957649231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,power_law_1.2,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,power_law_1.2,0.1792896032333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,power_law_1.2,0.1848896026611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,power_law_1.2,0.1918015956878662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,power_law_1.2,0.20448000431060792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,power_law_1.2,0.21541121006011962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,power_law_1.2,0.224019193649292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,power_law_1.2,0.2432703971862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,power_law_1.2,0.26318719387054446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,power_law_1.2,0.27825920581817626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,power_law_1.2,0.31100800037384035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,power_law_1.2,0.3876735925674438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,power_law_1.2,0.381113600730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,power_law_1.2,0.4822527885437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,power_law_1.2,0.5582143783569335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,power_law_1.2,0.7376128196716308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,power_law_1.2,0.927558422088623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,power_law_1.2,1.2591936111450195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,power_law_1.2,1.571782398223877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,power_law_1.2,2.2447488784790037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,power_law_1.2,2.8993215560913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,power_law_1.01,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,power_law_1.01,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,power_law_1.01,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,power_law_1.01,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,power_law_1.01,0.09189119935035706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,power_law_1.01,0.12827520370483397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,power_law_1.01,0.17439359426498413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,power_law_1.01,0.24142720699310302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,power_law_1.01,0.2944447994232178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,power_law_1.01,0.44637441635131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,power_law_1.01,0.526963186264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,balanced,0.03690666705369949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,balanced,0.038032000263532005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,balanced,0.03867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,balanced,0.04008533308903376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,balanced,0.039333333571751915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07452266911665599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.09144000212351482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.10966400305430095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.14546133081118265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.1720693310101827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.23653332392374674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.29501332839330036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.06038399934768677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.06698880195617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.07306240200996399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.12003200054168701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.144159996509552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.174892795085907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.2140415906906128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.28238720893859864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.3561023950576782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.501087999343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.6138944149017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,power_law_1.2,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,power_law_1.2,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,power_law_1.2,0.06972159743309021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,power_law_1.2,0.082777601480484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,power_law_1.2,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,power_law_1.2,0.10538239479064941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,power_law_1.2,0.10891519784927368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,power_law_1.2,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,power_law_1.2,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,power_law_1.2,0.11822079420089722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,power_law_1.2,0.1253119945526123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,power_law_1.2,0.13131519556045532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,power_law_1.2,0.13659520149230958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,power_law_1.2,0.14835840463638306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,power_law_1.2,0.1759104013442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,power_law_1.2,0.19575040340423583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,power_law_1.2,0.23632640838623048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,power_law_1.2,0.28158719539642335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,power_law_1.2,0.35564160346984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,power_law_1.2,0.437011194229126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,power_law_1.2,0.59170560836792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,power_law_1.2,0.7372672080993652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,power_law_1.2,1.0545408248901367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,power_law_1.2,1.3544639587402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,power_law_1.2,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,power_law_1.2,0.045459198951721194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,power_law_1.2,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,power_law_1.2,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,power_law_1.2,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,power_law_1.2,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,power_law_1.2,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,power_law_1.2,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,power_law_1.2,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,power_law_1.2,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,power_law_1.2,0.10184320211410522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,power_law_1.2,0.13468799591064454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,power_law_1.2,0.17375999689102173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,power_law_1.2,0.22100479602813722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,power_law_1.2,0.3152064085006714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,power_law_1.2,0.437011194229126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,power_law_1.2,0.6054656028747558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,power_law_1.2,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,power_law_1.2,0.19294079542160034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,power_law_1.2,0.22234880924224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,power_law_1.2,0.24340479373931884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,power_law_1.2,0.29944961071014403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,power_law_1.2,0.3354367971420288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,power_law_1.2,0.3420736074447632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,power_law_1.2,0.2714047908782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,power_law_1.2,0.2841408014297485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,power_law_1.2,0.26959359645843506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,power_law_1.2,0.285478401184082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,power_law_1.2,0.2844928026199341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,power_law_1.2,0.29356160163879397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,power_law_1.2,0.3377216100692749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,power_law_1.2,0.33523199558258054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,power_law_1.2,0.3721920013427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,power_law_1.2,0.3667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,power_law_1.2,0.4109055995941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,power_law_1.2,0.49909119606018065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,power_law_1.2,0.5112319946289062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,power_law_1.2,0.6136256217956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,power_law_1.2,0.726041603088379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,power_law_1.2,0.9116352081298829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,power_law_1.2,1.0566271781921386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,power_law_1.2,1.3005824089050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,power_law_1.2,1.715839958190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.11196160316467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.12803200483322144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.1543552041053772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.20924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.25546879768371583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.36056320667266845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.43476481437683107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,power_law_1.2,0.6703743934631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,power_law_1.2,0.7862016201019287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,balanced,0.03173866619666418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,balanced,0.033200000723203026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,balanced,0.03346666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,balanced,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,balanced,0.03629333277543386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,balanced,0.04035733391841253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,balanced,0.08878933389981587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,balanced,0.12223466237386067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,balanced,0.15384533007939658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,balanced,0.20825066169102988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,balanced,0.259770671526591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,balanced,0.05835733314355215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,balanced,0.08410132924715678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,balanced,0.11686933040618896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,balanced,0.11924266815185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,balanced,0.11815466483434041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,balanced,0.12061867117881775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,balanced,0.11895466844240825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,balanced,0.12155200044314067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,balanced,0.1223360002040863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,balanced,0.12482133507728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,balanced,0.12666666507720947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,balanced,0.1322986682256063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,balanced,0.13334932923316956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,balanced,0.13961066802342734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,balanced,0.1518880029519399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,balanced,0.15871466199556986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,balanced,0.18197333812713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,balanced,0.20279467105865479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,balanced,0.24819199244181314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,balanced,0.28666667143503827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,balanced,0.385045329729716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,balanced,0.4669119914372762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,balanced,0.6636373202006022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,balanced,0.8326880137125651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.08924800157546997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.09052799940109253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.12367360591888428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.13894399404525756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.1552575945854187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.134169602394104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.19631999731063843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.23675520420074464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.32605440616607667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,power_law_1.01,0.35477120876312257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,power_law_1.2,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,power_law_1.01,0.4546624183654785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,power_law_1.2,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,power_law_1.2,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,power_law_1.2,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,power_law_1.2,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,power_law_1.2,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,power_law_1.2,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,power_law_1.2,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,power_law_1.2,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,power_law_1.2,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,power_law_1.2,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.2,0.13486080169677733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.2,0.147705602645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.2,0.18733439445495606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.2,0.22629120349884033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.2,0.3180288076400757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.2,0.42055039405822753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.2,0.720255994796753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.2,0.8888128280639649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,power_law_1.01,0.6728384017944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,power_law_1.01,0.9337087631225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,power_law_1.01,1.3108991622924804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,power_law_1.01,2.0116352081298827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.07590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.08462079763412475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.09975039958953857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.12832640409469603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.16284159421920777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.18086400032043456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.24041600227355958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,balanced,0.043749332427978516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,balanced,0.061978667974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,balanced,0.09373333056767781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.31201279163360596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.45326719284057615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.5318463802337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,power_law_1.01,2.636012840270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.8333184242248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,power_law_1.2,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,power_law_1.2,0.10576000213623046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,power_law_1.2,0.11233919858932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,power_law_1.2,0.1451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,power_law_1.2,0.15091840028762818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,power_law_1.2,0.15218559503555298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,power_law_1.2,0.15756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,power_law_1.2,0.16230399608612062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,power_law_1.2,0.1736256003379822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,power_law_1.2,0.17469439506530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,power_law_1.2,0.18157440423965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,power_law_1.2,0.19243520498275757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,power_law_1.2,0.19850879907608032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,power_law_1.2,0.21171839237213136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,power_law_1.2,0.23788158893585204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,power_law_1.2,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,power_law_1.2,0.3034879922866821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,power_law_1.2,0.3343744039535522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,power_law_1.2,0.42748160362243653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,power_law_1.2,0.4830848217010498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,power_law_1.2,0.6570623874664306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,power_law_1.2,0.8387583732604981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,power_law_1.2,1.1402239799499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,power_law_1.2,1.5083328247070313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.0917759895324708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.11784960031509399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.17429759502410888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.19107199907302858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.23549439907073974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.2590912103652954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.3009536027908325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.35253119468688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.3620991945266724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.3379008054733276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.3583103895187378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.3605760097503662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.3434751987457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.39089920520782473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.3761919975280762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.3892416000366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.3879551887512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.4308159828186035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.4573376178741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.49185919761657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.5573760032653808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,0.6680511951446533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,0.7380288124084473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,0.9339903831481934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.1198464393615724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,1.5225088119506835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,1.915020751953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,balanced,0.028399998943010967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,balanced,0.027535999814669292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,balanced,0.02959466725587845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,balanced,0.04860800007979075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,balanced,0.04212800165017446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,balanced,0.04819199939568838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,balanced,0.06596800188223521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,balanced,0.0766186664501826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,balanced,0.09128000338872273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,balanced,0.11289599537849426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,balanced,0.13767466942469278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,balanced,0.16872533162434897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,balanced,0.22819733619689941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,balanced,0.281056006749471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,power_law_1.01,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,power_law_1.01,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,power_law_1.01,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,power_law_1.01,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.01,0.07002239823341369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.01,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.01,0.11868159770965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.01,0.15229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.08968960046768189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.12170239686965942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.01,0.2082751989364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.14160000085830687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.15824639797210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.24634239673614503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.32965760231018065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.4254335880279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.6285120010375976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.7927487850189209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.01,0.27940480709075927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.01,0.31410560607910154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,balanced,0.06381333371003468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,balanced,0.06528000036875407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,balanced,0.06394133468468984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,balanced,0.07597333192825317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,balanced,0.09184533357620239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,balanced,0.09590400258700053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,balanced,0.11762133240699768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,balanced,0.12877333164215088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,balanced,0.17046932379404703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,balanced,0.2039146622021993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,balanced,0.2772586743036906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,balanced,0.35304534435272217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,balanced,0.5003360112508138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,balanced,0.6373973290125529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,power_law_1.2,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,power_law_1.2,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,power_law_1.2,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,power_law_1.2,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,power_law_1.2,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,power_law_1.2,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,power_law_1.2,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,power_law_1.2,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,power_law_1.2,0.18537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,power_law_1.2,0.21287040710449218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,power_law_1.2,0.28873600959777834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,power_law_1.2,0.3935168027877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,power_law_1.2,0.49514241218566896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,power_law_1.2,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,power_law_1.2,0.020588800311088562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,power_law_1.2,0.02099840044975281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,power_law_1.2,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,power_law_1.2,0.02391040027141571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,power_law_1.2,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,power_law_1.2,0.0229312002658844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,power_law_1.2,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,power_law_1.2,0.027302399277687073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,power_law_1.2,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,power_law_1.2,0.028960001468658448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,power_law_1.2,0.02980479896068573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,power_law_1.2,0.030265599489212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,power_law_1.2,0.02919679880142212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,power_law_1.2,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,power_law_1.2,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.14469120502471924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,power_law_1.2,0.6503680229187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.17978880405426026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,balanced,0.0678773323694865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,balanced,0.10750933488210042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,balanced,0.12483732899030049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,balanced,0.12682666381200156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,balanced,0.12544533610343933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,balanced,0.1253546675046285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,balanced,0.1275200049082438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,balanced,0.1280586620171865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,balanced,0.12954666217168173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,balanced,0.13016000390052795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,balanced,0.13295466701189676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,balanced,0.13768532872200012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,balanced,0.13938132921854654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,balanced,0.172650674978892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,balanced,0.14871467153231302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,balanced,0.1844266653060913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,balanced,0.1742239991823832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,balanced,0.259770671526591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,balanced,0.24424533049265543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,balanced,0.3293226758639018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,balanced,0.34893866380055744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,balanced,0.5025546550750732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,balanced,0.5811253388722738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.08586239814758301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.1227903962135315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,balanced,0.036271999279658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,balanced,0.036746665835380554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,balanced,0.038319999972979225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,balanced,0.04063999901215235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,balanced,0.05540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,balanced,0.06459199885527293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,balanced,0.07262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,balanced,0.08269333342711131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,balanced,0.10149332880973816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,balanced,0.12811733285586038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,balanced,0.1325973371664683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,balanced,0.1910933256149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,balanced,0.23213332891464233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,balanced,0.3174239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,balanced,0.4065866470336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.15047039985656738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.18234879970550538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.223801589012146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,balanced,0.045909335215886436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,balanced,0.0528053343296051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,balanced,0.05451733370621999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,balanced,0.05810666580994924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,balanced,0.059658666451772056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,balanced,0.06459199885527293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,balanced,0.08539733290672302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,balanced,0.11109866698582967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,balanced,0.12158399820327759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,balanced,0.15499200423558554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,balanced,0.14724799990653992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,balanced,0.18653333187103271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,balanced,0.20121600230534872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,balanced,0.2669866681098938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,balanced,0.30456000566482544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.276307201385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.4335487842559814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,balanced,0.06283199787139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,balanced,0.06211733321348826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,balanced,0.062128002444903054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,balanced,0.06213866670926412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,balanced,0.06509866813818614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,balanced,0.06619200110435486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,balanced,0.06897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,balanced,0.07128533224264781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,balanced,0.07161066432793935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,balanced,0.07790400087833405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,balanced,0.08121599753697713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,balanced,0.08871466914812724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,balanced,0.10012267033259074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,balanced,0.11778133114178975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,balanced,0.1313813328742981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,balanced,0.16926399866739908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,balanced,0.20374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,balanced,0.2718186577161153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,balanced,0.3457706769307454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.544985580444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,balanced,0.03402133285999298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,balanced,0.036576000352700554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,balanced,0.03798400113979975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,balanced,0.037258667250474296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,balanced,0.05150400102138519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,balanced,0.0566293348868688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,balanced,0.07905066510041554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,balanced,0.08988799651463826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,balanced,0.1176479955514272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,balanced,0.13402666648228964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,balanced,0.18330132961273193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,balanced,0.2434719999631246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,balanced,0.3485706647237142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,balanced,0.46490665276845294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,balanced,0.05161066850026449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,balanced,0.05300800005594889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,balanced,0.05454933146635691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,balanced,0.05526400109132131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,balanced,0.054917335510253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,balanced,0.057333335280418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,balanced,0.05677333474159241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,balanced,0.06080000102519989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,balanced,0.06638933221499126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,balanced,0.07413866619269054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,balanced,0.07849066456158955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,balanced,0.09074667096138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,balanced,0.10749333103497823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,balanced,0.12356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,balanced,0.15476266543070474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,balanced,0.17577600479125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,balanced,0.23040000597635904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,balanced,0.2906506657600403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,balanced,0.4026666482289632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,balanced,0.494704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.12341760396957398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.12115199565887451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.1550976037979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.193612802028656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.2073215961456299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.2569792032241821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.14409600496292113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.16435199975967407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.16767359972000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.18744319677352905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.23545598983764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.24488320350646972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,power_law_1.01,0.33276801109313964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,power_law_1.01,0.3769536018371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,power_law_1.01,0.5879039764404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,power_law_1.01,0.6918079853057861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,power_law_1.01,0.8587967872619628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,power_law_1.01,1.1735487937927247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,power_law_1.01,2.0003839492797852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,power_law_1.01,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,power_law_1.01,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,power_law_1.01,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,power_law_1.01,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,power_law_1.01,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,power_law_1.01,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,power_law_1.01,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,power_law_1.01,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,power_law_1.01,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,power_law_1.01,0.09215360283851623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,power_law_1.01,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,power_law_1.01,0.13464959859848022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,power_law_1.01,0.17250560522079467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,power_law_1.01,0.20924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,power_law_1.01,0.29057281017303466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,power_law_1.01,0.3596735954284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,power_law_1.01,2.4563583374023437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,power_law_1.01,3.7587265014648437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,power_law_1.01,4.273350524902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,power_law_1.2,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,power_law_1.2,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,power_law_1.2,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,power_law_1.2,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,power_law_1.2,0.09819520115852357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,power_law_1.2,0.10573439598083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,power_law_1.2,0.10852479934692383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,power_law_1.2,0.11122560501098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,power_law_1.2,0.11522560119628907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,power_law_1.2,0.11434880495071412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,power_law_1.2,0.12354559898376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,power_law_1.2,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,power_law_1.2,0.13973759412765502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,power_law_1.2,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,power_law_1.2,0.1606335997581482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,power_law_1.2,0.17385599613189698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,power_law_1.2,0.2091775894165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,power_law_1.2,0.22999041080474852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,power_law_1.2,0.2951616048812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,power_law_1.2,0.3400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,power_law_1.2,0.46582398414611814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,power_law_1.2,0.5827392101287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,power_law_1.2,0.781766414642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,power_law_1.2,1.0948800086975097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.14699519872665406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.23608319759368895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.3861824035644531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.5713664054870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.7660736083984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.134540843963623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.1975744247436524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.2638655662536622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.2652992248535155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.3406335830688476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.3875840187072754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.4104000091552735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.451193618774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.5079104423522949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.5394944190979003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.6189504623413087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,power_law_1.2,1.773811149597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,power_law_1.2,1.7923776626586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.01343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.025555229187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,power_law_1.2,2.3077247619628904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,power_law_1.2,2.6225664138793947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,power_law_1.2,3.1649023056030274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,power_law_1.2,3.537004852294922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,power_law_1.2,4.53625602722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,power_law_1.2,5.635603332519532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,power_law_1.01,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,power_law_1.01,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,power_law_1.01,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,power_law_1.01,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,power_law_1.01,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,power_law_1.01,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,power_law_1.01,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,power_law_1.01,0.07267199754714966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,power_law_1.01,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,power_law_1.01,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,power_law_1.01,0.14110080003738404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,power_law_1.01,0.16504960060119628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,power_law_1.01,0.21373438835144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,power_law_1.01,0.2656768083572388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,power_law_1.01,0.3758336067199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,power_law_1.01,0.44257278442382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,power_law_1.01,0.6960192203521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,power_law_1.01,0.8958399772644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,balanced,0.0702453354994456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,balanced,0.09570133686065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,balanced,0.14479999740918478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,balanced,0.24633065859476724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,balanced,0.4440213441848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,balanced,0.8413440386454264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,balanced,0.8447573184967041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,balanced,0.8487573464711508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,balanced,0.8555413087209066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,balanced,0.857205311457316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,balanced,0.8600800037384033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,balanced,0.8636159896850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,balanced,0.8774186770121256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,balanced,0.8839840094248453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,balanced,0.8967253367106119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,balanced,0.9074772993723551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,balanced,0.9297706286112467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,balanced,0.978437344233195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,balanced,1.0100959936777751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,balanced,1.1125226815541585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,balanced,1.184666633605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,balanced,1.3316319783528645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,balanced,1.4971572558085124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,balanced,1.9984532992045085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,balanced,2.236037254333496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,balanced,3.1998348236083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,balanced,3.952773412068685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,power_law_1.2,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,power_law_1.2,0.11491199731826782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,power_law_1.2,0.20213758945465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,power_law_1.2,0.20622079372406005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,power_law_1.2,0.2078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,power_law_1.2,0.21022720336914064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,power_law_1.2,0.21747839450836182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,power_law_1.2,0.21987199783325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,power_law_1.2,0.21751680374145507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,power_law_1.2,0.2197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,power_law_1.2,0.22712960243225097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,power_law_1.2,0.23235199451446534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,power_law_1.2,0.25001599788665774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,power_law_1.2,0.2716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,power_law_1.2,0.2863231897354126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,power_law_1.2,0.2832256078720093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,power_law_1.2,0.3244096040725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,power_law_1.2,0.37268478870391847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,power_law_1.2,0.45593600273132323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,power_law_1.2,0.5450496196746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,power_law_1.2,0.7205887794494629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.10947200059890747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.15707520246505738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.1705024003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.225382399559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.28072960376739503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.4252799987792969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.5334464073181152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,power_law_1.2,0.9005951881408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,power_law_1.2,1.2438143730163573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,power_law_1.2,1.5859071731567382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.10100480318069457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.10549119710922242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.12940800189971924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.15857919454574584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.18458240032196044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.22586240768432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.2831104040145874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.3804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.44376320838928224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.6155968189239502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.8884415626525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,power_law_1.2,2.346668815612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,power_law_1.2,3.0712575912475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,power_law_1.2,0.028620800375938414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,power_law_1.2,0.028191998600959778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,power_law_1.2,0.027750399708747864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,power_law_1.2,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,power_law_1.2,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,power_law_1.2,0.031455999612808226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,power_law_1.2,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,power_law_1.2,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,power_law_1.2,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,power_law_1.2,0.16479359865188598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,power_law_1.2,0.2225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,power_law_1.2,0.2545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.09221760034561158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.09735040068626404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.10263680219650269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.10527360439300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.1059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.1108288049697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.11541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.1194111943244934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.12673280239105225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.13765759468078614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.13702399730682374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.15226880311965943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.1672063946723938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.20768001079559326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.23575680255889891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.2996927976608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.3972991943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.4892223834991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.659219217300415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,0.9168576240539551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,power_law_1.2,0.3949631929397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.1766719818115234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,power_law_1.2,0.48078079223632814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,balanced,0.07648533085982005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,balanced,0.08872532844543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,balanced,0.12103466192881267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,balanced,0.174234668413798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,balanced,0.2660106619199117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,balanced,0.27455999453862506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,balanced,0.27165865898132324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,balanced,0.2743733326594035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,balanced,0.2741706569989522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,balanced,0.28034667174021405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,balanced,0.2868480086326599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,balanced,0.2837653358777364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,balanced,0.29480000336964923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,balanced,0.30018667380015057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,balanced,0.3069760004679362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,balanced,0.315557340780894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,balanced,0.34228265285491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,balanced,0.3693173329035441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,balanced,0.4278879960378011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,balanced,0.487557331720988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,balanced,0.6000959873199463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,balanced,0.7115360101064047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,balanced,0.9961439768473307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,balanced,1.204581340154012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,balanced,1.7543840408325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,balanced,2.264677365620931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.07511680126190186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.08301439881324768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.1260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.15587199926376344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.17221120595932007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.2194751977920532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.26604158878326417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.3492799997329712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.5333055973052978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,0.781113576889038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.0329792022705078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,1.6427711486816405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,2.1873023986816404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,balanced,0.0420959989229838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,balanced,0.04539200166861216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,balanced,0.06173333525657654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,balanced,0.08153066535790761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,balanced,0.1169599990049998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,balanced,0.14171199997266135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,balanced,0.19338132937749228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,balanced,0.23891200621922812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,balanced,0.33713066577911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,balanced,0.41840000947316486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.10510079860687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.1835263967514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.21755518913269042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.23432319164276122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.23583359718322755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.2855999946594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.28485119342803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.3178303956985474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.3325056076049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.2831360101699829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.28784639835357667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.2881983995437622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.2964672088623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.32648959159851076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.3442368030548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.3814336061477661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,power_law_1.2,0.4040832042694092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,power_law_1.2,0.4645247936248779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,power_law_1.2,0.529856014251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,power_law_1.2,0.6737664222717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,power_law_1.2,0.9698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,power_law_1.2,1.276038360595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.04200960099697113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.11213439702987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.1347391963005066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.18462079763412476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.21807360649108887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.3162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.4181824207305908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,power_law_1.2,1.4496447563171386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.07692800164222717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.08094720244407654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.09193599820137024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.10220160484313964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.10984959602355956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.1265663981437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.14606080055236817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.18535679578781128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.20239360332489015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.25107200145721437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.34163200855255127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,power_law_1.2,2.0385343551635744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.4573376178741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.5237823963165283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.06090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.11100159883499146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.14072959423065184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.1797376036643982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.23596160411834716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.32410240173339844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.47839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.5642560005187989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,power_law_1.2,3.081267166137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,power_law_1.2,0.845792007446289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.0115519523620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,power_law_1.2,4.192723083496094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,power_law_1.01,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,power_law_1.01,0.10019840002059936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,power_law_1.01,0.2476288080215454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,power_law_1.01,0.3363647937774658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,power_law_1.01,0.4508224010467529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,power_law_1.01,0.6211520195007324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,power_law_1.01,0.6448832035064698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,power_law_1.01,0.6845056056976319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,power_law_1.01,0.6892096042633057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,power_law_1.01,0.7062655925750733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,power_law_1.01,0.7284992218017579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,power_law_1.01,0.7453120231628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,power_law_1.01,0.7755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,power_law_1.01,0.79683837890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,power_law_1.01,0.817024040222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,power_law_1.01,0.860102367401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,power_law_1.01,0.9168895721435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,power_law_1.01,0.9754688262939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,power_law_1.01,1.1448320388793944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,power_law_1.01,1.1754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,power_law_1.01,1.3853695869445801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,power_law_1.01,1.631648063659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,power_law_1.01,2.0872896194458006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,power_law_1.01,2.504921531677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,power_law_1.01,3.36511344909668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,power_law_1.01,4.290617752075195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,power_law_1.2,0.13217920064926147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,power_law_1.2,0.1496448040008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,power_law_1.2,0.165664005279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,power_law_1.2,0.19914239645004272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,power_law_1.2,0.2182015895843506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,power_law_1.2,0.2081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,power_law_1.2,0.21008639335632323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,power_law_1.2,0.22405118942260743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,power_law_1.2,0.22428159713745116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,power_law_1.2,0.22648959159851073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,power_law_1.2,0.2388672113418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,2,power_law_1.2,0.23546240329742432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,2,power_law_1.2,0.24540159702301026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,2,power_law_1.2,0.2564160108566284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,2,power_law_1.2,0.2820224046707153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.2,0.29258880615234373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.2,0.34369280338287356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.2,0.382860803604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.2,0.48183040618896483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.2,0.5782015800476075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.2,0.769708776473999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.2,0.9838144302368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.2,1.4384896278381347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.2,1.7697280883789062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.05692800283432007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.10917760133743286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.1362048029899597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.17728639841079713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.21600000858306884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,balanced,0.0399893323580424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,balanced,0.054330666859944664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,balanced,0.08330666522185008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,balanced,0.08775466680526733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,balanced,0.08809066812197368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,balanced,0.09053333600362141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,balanced,0.09196266531944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,balanced,0.08945066730181377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,balanced,0.0909546713034312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,balanced,0.0930560032526652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,balanced,0.0936906635761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,balanced,0.09531733393669128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,balanced,0.10313066840171814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,balanced,0.1030346651871999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,balanced,0.10759466886520386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,balanced,0.12247467041015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,balanced,0.1630293329556783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,balanced,0.17709332704544067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,balanced,0.24055467049280801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,balanced,0.25933865706125897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,balanced,0.3811200062433879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,balanced,0.45495466391245526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,balanced,0.6222666501998901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,balanced,0.7981119950612386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.2993216037750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.3944704055786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.17063039541244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.21369600296020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.3003904104232788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.34965760707855226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.13841279745101928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.13969279527664186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.142739200592041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.14539519548416138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.134879994392395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.13923200368881225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.13078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.12320640087127685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.12943359613418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.14209920167922974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.14624639749526977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.1521407961845398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.1792255997657776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.1699136018753052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.21133439540863036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.21500799655914307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.26440958976745604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.251475191116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.31893119812011717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.3710207939147949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,power_law_1.2,0.49864959716796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,power_law_1.2,0.6395584106445312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,power_law_1.2,0.9055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,power_law_1.2,1.0743424415588378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,power_law_1.01,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,power_law_1.01,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,power_law_1.01,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,power_law_1.01,0.04953599870204926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,power_law_1.01,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,power_law_1.01,0.08019840121269226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,power_law_1.01,0.10230400562286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,power_law_1.01,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,power_law_1.01,0.1424831986427307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,power_law_1.01,0.18266240358352662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,power_law_1.01,0.21649279594421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,power_law_1.01,0.3028671979904175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,power_law_1.01,0.36345601081848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,power_law_1.01,0.5331264019012452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,power_law_1.01,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,power_law_1.01,0.7036928176879883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,power_law_1.01,0.10905599594116211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,power_law_1.01,0.1071936011314392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,power_law_1.01,0.10855040550231934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,power_law_1.01,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,power_law_1.01,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,power_law_1.01,0.0972927987575531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,power_law_1.01,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,power_law_1.01,0.10370559692382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,power_law_1.01,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,power_law_1.01,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,power_law_1.01,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,power_law_1.01,0.10639359951019287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,power_law_1.01,0.10891519784927368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,power_law_1.01,0.11969280242919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,power_law_1.01,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,power_law_1.01,0.12725759744644166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,power_law_1.01,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,power_law_1.01,0.1482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,power_law_1.01,0.17121280431747438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,power_law_1.01,0.2052608013153076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,power_law_1.01,0.2573888063430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,power_law_1.01,0.30687999725341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,power_law_1.01,0.4243840217590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,power_law_1.01,0.544320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,power_law_1.01,0.7335423946380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,power_law_1.01,0.9557184219360352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.01,0.0889024019241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.01,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.01,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.01,0.08721280097961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.01,0.08792319893836975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.01,0.09012479782104492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.01,0.09254400134086609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.01,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.01,0.09492480158805847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.01,0.09697279930114747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.01,0.09365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.01,0.10047999620437623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.01,0.10554879903793335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.01,0.11439360380172729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.01,0.12903039455413817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.01,0.17239680290222167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.01,0.19634560346603394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.01,0.2546367883682251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.01,0.2756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.01,0.383020806312561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.01,0.46613759994506837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.01,0.6189375877380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.01,0.8056256294250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,balanced,0.03804266701141993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,balanced,0.04337066908677419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,balanced,0.046181331078211464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,balanced,0.05089599887530009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,balanced,0.05031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,balanced,0.053173333406448364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,balanced,0.062122667829195656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,balanced,0.06154666841030121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,balanced,0.07696533203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,balanced,0.09111467003822327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,balanced,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,balanced,0.12665067116419473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,balanced,0.1702773372332255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,balanced,0.2039466698964437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,balanced,0.29231466849644977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,balanced,0.3999040126800537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,balanced,0.5849119822184244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,balanced,0.7568213144938151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,power_law_1.01,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,power_law_1.01,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,power_law_1.01,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,power_law_1.01,0.03203839957714081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,power_law_1.01,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,power_law_1.01,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,power_law_1.01,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,power_law_1.01,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,power_law_1.01,0.05041279792785645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,power_law_1.01,0.09039360284805298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,power_law_1.01,0.07805439829826355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,power_law_1.01,0.09949439764022827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,power_law_1.01,0.1140671968460083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,power_law_1.01,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,power_law_1.01,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,power_law_1.01,0.1315840005874634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,power_law_1.01,0.13668479919433593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,power_law_1.01,0.14967039823532105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,power_law_1.01,0.18174079656600953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,power_law_1.01,0.1485759973526001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,power_law_1.01,0.20829439163208008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,power_law_1.01,0.19219839572906494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,power_law_1.01,0.2687551975250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,power_law_1.01,0.25399041175842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,power_law_1.01,0.323686408996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,power_law_1.01,0.31931519508361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,power_law_1.01,0.4403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,power_law_1.01,0.5542912006378173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,power_law_1.01,0.776416015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,power_law_1.01,0.9938816070556641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,power_law_1.01,1.431065559387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,power_law_1.01,1.8492351531982423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.06873599886894226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.08922880291938781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.09162240028381348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.0908415973186493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.10232959985733033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.11274240016937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.11689599752426147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.12933119535446166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.14979200363159179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.17529599666595458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.21412479877471924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.24087679386138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.30869760513305666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.37947518825531007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.17970559597015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.2325119972229004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.3555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.3856832027435303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.3977855920791626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.40729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.416428804397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.42855038642883303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.4413887977600098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.45500798225402833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.4703104019165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.01,0.4805823802947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.01,0.5048831939697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.01,0.5380224227905274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.01,0.6170752048492432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.01,0.6761983871459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.01,0.8298239707946777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.021401596069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.01,1.3077568054199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.01,1.6072000503540038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.01,2.1691455841064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.01,2.5963775634765627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.01,3.7047168731689455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.01,4.883993530273438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.5276031970977784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.2,0.09728000164031983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.2,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.2,0.11572480201721191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.2,0.12278399467468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.2,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.2,0.11515519618988038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.2,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.2,0.12171519994735717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.2,0.12039680480957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.2,0.12200959920883178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.2,0.12767360210418702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.2,0.12655999660491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.2,0.13198720216751098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.2,0.13833600282669067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.2,0.14527360200881959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.2,0.15537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.2,0.17890559434890746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.2,0.19415680170059205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.2,0.24134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.2,0.2817408084869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.2,0.35571839809417727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.2,0.39643518924713134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.2,0.5423488140106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.2,0.6515967845916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.2,1.0238783836364747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.2,1.2913599967956544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.6524735927581787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,power_law_1.2,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,power_law_1.2,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,power_law_1.2,0.033939200639724734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,power_law_1.2,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,power_law_1.2,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,power_law_1.2,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,power_law_1.2,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,power_law_1.2,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,power_law_1.2,0.06704639792442321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,power_law_1.2,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,power_law_1.2,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,power_law_1.2,0.11683839559555054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,power_law_1.2,0.15228159427642823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,power_law_1.2,0.1970944046974182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,power_law_1.2,0.26328320503234864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,power_law_1.2,0.3114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,0.9345151901245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,power_law_1.2,0.49707517623901365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,power_law_1.2,0.5285312175750733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,balanced,0.03181866556406021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,balanced,0.03596800069014231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,balanced,0.058890665570894875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,balanced,0.05972800155480703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,balanced,0.06085333228111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,balanced,0.06251200040181477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,balanced,0.06579733391602834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,balanced,0.07277333239714305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,balanced,0.0730506678422292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,balanced,0.07981866598129272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,balanced,0.08732266227404277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,balanced,0.10036266843477885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,balanced,0.11105066537857056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,balanced,0.13754133383433023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,balanced,0.16988267501195273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,balanced,0.2230186661084493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,balanced,0.26735466718673706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,balanced,0.3694186607996623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,balanced,0.4742293357849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.2080320358276366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,balanced,0.05562133093674978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,balanced,0.08984532952308655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,balanced,0.1362559994061788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,balanced,0.20988265673319498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,balanced,0.21101333697636923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,balanced,0.2145813306172689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,balanced,0.21565866470336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,balanced,0.21701333920160928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,balanced,0.2180639902750651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,balanced,0.22500799099604288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,balanced,0.2259733279546102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,balanced,0.22990399599075317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,balanced,0.23805334170659384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,balanced,0.2425439953804016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,balanced,0.25643734137217206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,balanced,0.29396265745162964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,balanced,0.318725327650706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,balanced,0.383786678314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,balanced,0.4591840108235677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,balanced,0.6363946596781412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,balanced,0.7581653594970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,balanced,1.064581314722697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,balanced,1.3036213715871174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,balanced,1.9418932596842449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,balanced,2.519770622253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.2,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.2,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.2,0.05354239940643311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.2,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.2,0.06039680242538452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.2,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.2,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.2,0.09402239918708802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.2,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.2,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.2,0.14810880422592163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.2,0.1691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.2,0.2145792007446289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.2,0.2359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.2,0.3231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.2,0.41296000480651857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.2,0.54999680519104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.2,0.6952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,balanced,0.052255998055140175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,balanced,0.05323199927806854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,balanced,0.08733866612116496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,balanced,0.12013333042462666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,balanced,0.1509119967619578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,balanced,0.15226667126019797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,balanced,0.1527839998404185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,balanced,0.15498133500417074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,balanced,0.15521066387494406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,balanced,0.15689067045847574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,balanced,0.15756799777348837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,balanced,0.16061333815256754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,balanced,0.1662399967511495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,balanced,0.16781866550445557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,balanced,0.1779359976450602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,balanced,0.18666666746139526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.19794134298960367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.21940267086029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.24278400341669717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.2885813315709432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.33220799763997394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.4139839808146159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.5212159951527914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.686954657236735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,balanced,0.895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.02905600070953369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.02919679880142212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.02968960106372833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.05315200090408325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.09768319725990296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.16181119680404663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.01,0.09857919812202454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.01,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.01,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.01,0.10257279872894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.2066943883895874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.01,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.01,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.01,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.01,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.01,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.01,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.01,0.10425599813461303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.01,0.10510079860687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.01,0.10824320316314698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.01,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.01,0.12121599912643433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.01,0.13679360151290892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.01,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.01,0.15491199493408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.01,0.18781440258026122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.01,0.21847679615020751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.01,0.29134719371795653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.01,0.27251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.01,0.3426815986633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,balanced,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.01,0.44652161598205564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,balanced,0.05840000013510386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,balanced,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,balanced,0.08149866759777069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.01,0.5987648010253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,balanced,0.07699733475844066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,balanced,0.07874666651089986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,balanced,0.08108800152937572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,balanced,0.07989866534868877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,balanced,0.08137066662311554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,balanced,0.08683733145395915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.01,0.7029312133789063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,balanced,0.08756800492604573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,balanced,0.09360532959302266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,balanced,0.09710933764775594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,balanced,0.10389866431554158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,balanced,0.11317867040634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,balanced,0.1288693348566691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,balanced,0.14416533708572388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,balanced,0.1785866618156433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,balanced,0.2072426676750183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,balanced,0.2714186708132426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,balanced,0.3323733409245809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.2692863941192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,power_law_1.2,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,power_law_1.2,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,power_law_1.2,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,power_law_1.2,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,power_law_1.2,0.08302720189094544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,power_law_1.2,0.11869440078735352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,power_law_1.2,0.14633599519729615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,power_law_1.2,0.19281280040740967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,power_law_1.2,0.23917438983917236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,power_law_1.2,0.3479423999786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,power_law_1.2,0.4670400142669678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.3046528100967407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,power_law_1.2,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,power_law_1.2,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,power_law_1.2,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,power_law_1.2,0.045203199982643126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,power_law_1.2,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,power_law_1.2,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,power_law_1.2,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,power_law_1.2,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,power_law_1.2,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.2,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.2,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.2,0.12097280025482178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.2,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.2,0.20266239643096923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,power_law_1.2,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.2,0.24760959148406983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.2,0.32945919036865234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.2,0.4289408206939697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,power_law_1.2,0.09777920246124268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,power_law_1.2,0.09972479939460754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,power_law_1.2,0.1001855969429016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,power_law_1.2,0.10744320154190064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,power_law_1.2,0.10695040225982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,power_law_1.2,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,power_law_1.2,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,power_law_1.2,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,power_law_1.2,0.12149759531021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,power_law_1.2,0.12394880056381226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,power_law_1.2,0.12892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,power_law_1.2,0.1438976049423218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,power_law_1.2,0.15946240425109864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,power_law_1.2,0.17951359748840331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,power_law_1.2,0.19251840114593505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,power_law_1.2,0.2526655912399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,power_law_1.2,0.2867072105407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,power_law_1.2,0.3821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,power_law_1.2,0.518016004562378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,power_law_1.2,0.6803135871887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,power_law_1.2,0.8867327690124511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,balanced,0.08418666323026021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,balanced,0.08386133114496867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,balanced,0.08500799536705017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,balanced,0.08629332979520161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,balanced,0.08845866719881694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,balanced,0.0972106655438741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,balanced,0.0974026620388031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,balanced,0.09687999884287517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,balanced,0.09436266620953877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,balanced,0.09880000352859497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,balanced,0.10604799787203471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,balanced,0.11111467083295186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,balanced,0.10244266192118327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,balanced,0.10533333818117778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,balanced,0.10884267091751099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,balanced,0.1122773289680481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,balanced,0.12377599875132243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,balanced,0.11959999799728394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,balanced,0.12821867068608603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,balanced,0.14985600113868713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,balanced,0.16858667135238647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,balanced,0.21402132511138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,balanced,0.2485226591428121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,balanced,0.34534398714701336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,balanced,0.4156586726506551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.06407679915428162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.0719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.0785215973854065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.0917568027973175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.10818560123443603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.13308800458908082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.15582720041275025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.20283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.23759360313415528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.3349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.44620161056518554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,power_law_1.2,0.6401088237762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,power_law_1.2,0.8287167549133301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.06632320284843445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.06353920102119445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.06687359809875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.11337599754333497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.13360639810562133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.16106879711151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.19854079484939574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.29497599601745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.3376895904541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.48487038612365724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6550079822540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,balanced,0.027776000400384266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,balanced,0.0272533322374026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,balanced,0.027306665976842243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,balanced,0.029365333418051403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,balanced,0.029306667546431225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,balanced,0.03178666780392329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,balanced,0.03322133421897888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,balanced,0.035045333206653595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,balanced,0.03889599939187368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,balanced,0.06608533362547557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,balanced,0.07421333094437917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,balanced,0.11522666613260905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,balanced,0.15641066431999207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,balanced,0.19915199279785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,power_law_1.2,0.08035839796066284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,power_law_1.2,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,power_law_1.2,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,power_law_1.2,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,power_law_1.2,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,power_law_1.2,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,power_law_1.2,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,power_law_1.2,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,power_law_1.2,0.08022400140762329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,power_law_1.2,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,power_law_1.2,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,power_law_1.2,0.10313600301742554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,power_law_1.2,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,power_law_1.2,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,power_law_1.2,0.15248639583587648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,power_law_1.2,0.1824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,power_law_1.2,0.22707839012145997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,power_law_1.2,0.28538880348205564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,power_law_1.2,0.4054975986480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,power_law_1.2,0.5138815879821778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,power_law_1.2,0.7374080181121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,power_law_1.2,0.9685376167297364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,power_law_1.01,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,power_law_1.01,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,balanced,0.06896533568700154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,balanced,0.08649067083994548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,balanced,0.09275733431180318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,balanced,0.09340266386667888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,balanced,0.09622400005658467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,balanced,0.09377066294352214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,power_law_1.01,0.06094719767570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,balanced,0.09615466992060344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,balanced,0.09633066256841023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,balanced,0.09709866841634114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,balanced,0.09870933492978413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,balanced,0.09880533814430237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,balanced,0.10584533214569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,balanced,0.10759466886520386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,power_law_1.01,0.06353279948234558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,balanced,0.10977066556612651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,balanced,0.11709333459536235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,balanced,0.12264532844225566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,balanced,0.1359999974568685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,power_law_1.01,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,balanced,0.1462506651878357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,balanced,0.19432000319163004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,balanced,0.2144213318824768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,balanced,0.28568534056345624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,power_law_1.01,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,balanced,0.36724801858266193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,power_law_1.01,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,balanced,0.4857120116551717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,power_law_1.01,0.1015552043914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,balanced,0.6452746788660685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,power_law_1.01,0.13731839656829833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,power_law_1.01,0.14899200201034546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,power_law_1.01,0.18803839683532714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,power_law_1.01,0.23763840198516845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,power_law_1.01,0.3179903984069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,power_law_1.01,0.4190847873687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,power_law_1.2,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,power_law_1.2,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,power_law_1.2,0.04472959935665131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,power_law_1.2,0.04476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,power_law_1.2,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,power_law_1.2,0.06366080045700073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,power_law_1.2,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,power_law_1.2,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,power_law_1.2,0.08536959886550903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,power_law_1.2,0.09663360118865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,power_law_1.2,0.13242239952087403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,power_law_1.2,0.16867200136184693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,power_law_1.2,0.22312960624694825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,power_law_1.2,0.27482879161834717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,power_law_1.2,0.4305471897125244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,power_law_1.2,0.5047743797302247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.07229440212249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.08661760091781616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.09432960152626038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.12986880540847778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.13589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.18880640268325805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.21524479389190673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.3177792072296143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4283455848693848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.6055935859680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.751577615737915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,balanced,0.08459200461705525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,balanced,0.09514133135477702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,balanced,0.12152000268300374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,balanced,0.1824586590131124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,balanced,0.27160000801086426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,balanced,0.27560534079869586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,balanced,0.27703466018040973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,balanced,0.2773600021998088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,balanced,0.27801599105199176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,balanced,0.2837653358777364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,balanced,0.2897226611773173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,balanced,0.292410671710968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,balanced,0.3001120090484619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,balanced,0.30686400334040326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,balanced,0.31355200211207074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,balanced,0.3280479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,balanced,0.36897067228953045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.4014773368835449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.47775999704996747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.5421013434727987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.7015466690063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.8605919679005941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,balanced,1.2597653071085613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,balanced,1.5944587389628093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,balanced,2.314319928487142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,balanced,3.0529120763142905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,power_law_1.2,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,power_law_1.2,0.1269376039505005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,power_law_1.2,0.17768319845199584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,power_law_1.2,0.24634881019592286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,power_law_1.2,0.2614912033081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,power_law_1.2,0.4505216121673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,power_law_1.2,0.5060927867889404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,power_law_1.2,0.5014592170715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,power_law_1.2,0.509280014038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,power_law_1.2,0.5483583927154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,power_law_1.2,0.5411712169647217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,power_law_1.2,0.5705408096313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,power_law_1.2,0.5545536041259765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,power_law_1.2,0.594758415222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,power_law_1.2,0.5949759960174561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,power_law_1.2,0.6129151821136475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,power_law_1.2,0.6777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,power_law_1.2,0.7628352165222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,power_law_1.2,0.7604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,power_law_1.2,0.8695103645324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,power_law_1.2,0.9658687591552735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,power_law_1.2,1.2209535598754884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,power_law_1.2,1.2271295547485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,power_law_1.2,1.6011455535888672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,power_law_1.2,2.1251199722290037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,power_law_1.01,0.07465599775314331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,power_law_1.01,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,power_law_1.01,0.13367040157318116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,power_law_1.01,0.1320255994796753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,power_law_1.01,0.14657280445098878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,power_law_1.01,0.1458176016807556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,power_law_1.01,0.14286719560623168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,power_law_1.01,0.15081599950790406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,power_law_1.01,0.15450880527496338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,power_law_1.01,0.1606592059135437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,power_law_1.01,0.16279679536819458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,power_law_1.01,0.16311039924621581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,power_law_1.01,0.18454400300979615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,power_law_1.01,0.19440000057220458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,power_law_1.01,0.20694398880004883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,power_law_1.01,0.24320640563964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,power_law_1.01,0.25107200145721437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,power_law_1.01,0.29202558994293215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,power_law_1.01,0.3737215995788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,power_law_1.2,2.846873664855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,power_law_1.01,0.4385791778564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,power_law_1.01,0.5590335845947265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,power_law_1.2,3.291980743408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,power_law_1.01,0.7332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,power_law_1.01,1.0748031616210938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,power_law_1.01,1.4783231735229492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,power_law_1.01,2.0388479232788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,power_law_1.2,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,power_law_1.2,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,power_law_1.2,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,power_law_1.2,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,power_law_1.2,0.1082751989364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,power_law_1.2,0.12540160417556762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,power_law_1.2,0.18376959562301637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,power_law_1.2,0.2241215944290161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,power_law_1.01,2.4747455596923826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,power_law_1.2,0.36186239719390867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,balanced,0.06858666737874348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,balanced,0.08784533540407817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,balanced,0.10824533303578694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,balanced,0.11174933115641277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,balanced,0.10903466741243999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,balanced,0.11098666985829671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,balanced,0.11210133632024129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,balanced,0.11365333199501038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,balanced,0.11392533779144287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,balanced,0.11719466249148051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,balanced,0.12120532989501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,balanced,0.1320799986521403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,balanced,0.13788800438245138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,balanced,0.1481226682662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,balanced,0.17101333538691202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,balanced,0.19181867440541586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,balanced,0.23037334283192953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,balanced,0.2730453411738078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,balanced,0.35438934961954754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,balanced,0.4505973259607951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,balanced,0.6108800172805786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,balanced,0.7876853148142496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,power_law_1.2,0.347705602645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.01,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.01,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.01,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.01,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.01,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.01,0.06656640172004699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.01,0.0680191993713379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.01,0.06774399876594543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.01,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.01,0.07641599774360656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.01,0.09237759709358215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.01,0.11064319610595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.01,0.13651200532913207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.01,0.16239999532699584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.01,0.21452159881591798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.01,0.2291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.01,0.30404479503631593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.01,0.37845759391784667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.01,0.5245312213897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.01,0.6867392063140869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,balanced,0.11714667081832886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,balanced,0.2079520026842753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,balanced,0.38050134976704914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,balanced,0.37666134039560956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,balanced,0.3778293530146281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,balanced,0.3793333371480306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,balanced,0.3813759883244832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,balanced,0.3824426730473836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,balanced,0.38142398993174237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,balanced,0.3829919894536336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,balanced,0.39469865957895917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,balanced,0.3908533255259196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,balanced,0.3944480021794637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,balanced,0.40724265575408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,balanced,0.43239466349283856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,balanced,0.4385013182957967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,balanced,0.46504000822703045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,balanced,0.5036906798680624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,balanced,0.5472053289413452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,balanced,0.6774133046468099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,balanced,0.7420586744944254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,balanced,0.9922026793162028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,balanced,1.261888027191162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,balanced,1.8438879648844402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,balanced,2.4091307322184243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,balanced,3.568805376688639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,balanced,4.793082555135091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.039027199149131775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.0827455997467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.09081599712371827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.10366079807281495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.11048320531845093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.1174847960472107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.12376320362091064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.14370559453964232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.16992640495300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.20885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.25754880905151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.35253760814666746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.41546878814697263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.5822144031524659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.0605504035949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.029708799719810487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.3997247695922852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.05161600112915039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.0883072018623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.12673280239105225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.1702463984489441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,power_law_1.2,0.24179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,power_law_1.2,0.270630407333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,power_law_1.2,0.41832962036132815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,power_law_1.01,0.07492480278015137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,power_law_1.01,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,power_law_1.2,0.5019584178924561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,power_law_1.01,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,power_law_1.01,0.1579584002494812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,power_law_1.01,0.1670016050338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,power_law_1.01,0.26334080696105955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,power_law_1.01,0.27155840396881104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,power_law_1.01,0.27913599014282225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,power_law_1.01,0.287225604057312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,power_law_1.01,0.2888767957687378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,power_law_1.01,0.2996351957321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,power_law_1.01,0.30013439655303953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,power_law_1.01,0.31014399528503417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,power_law_1.01,0.3264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,power_law_1.01,0.3344383955001831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,power_law_1.01,0.3495935916900635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,power_law_1.01,0.35224959850311277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,power_law_1.01,0.3963776111602783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,power_law_1.01,0.4284736156463623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,power_law_1.01,0.5034495830535889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,power_law_1.01,0.5806848049163819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,power_law_1.01,0.7302976131439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,power_law_1.01,0.7875455856323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,power_law_1.01,1.0583488464355468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,power_law_1.01,1.3270400047302247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,power_law_1.01,1.8514816284179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,power_law_1.01,2.2688575744628907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,balanced,0.06940266489982605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,balanced,0.10143466790517171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,balanced,0.15557333827018738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,balanced,0.25121599435806274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,balanced,0.3505866527557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,balanced,0.35068265597025555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,balanced,0.3569386800130208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,balanced,0.3566933472951253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,balanced,0.3644213279088338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,balanced,0.3712693452835083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,balanced,0.3734346628189087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,balanced,0.3741066853205363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,balanced,0.38913599650065106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,balanced,0.3864479859670003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,balanced,0.4041759967803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,balanced,0.4262239933013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,balanced,0.4455893437067668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,balanced,0.49137600262959796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,balanced,0.5291839838027954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,balanced,0.6119626760482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,balanced,0.7154293060302734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,balanced,0.8947199980417887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,balanced,1.1253493626912434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,balanced,1.4462933540344238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,balanced,1.9196160634358723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.01,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.01,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.01,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.01,0.11877119541168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.01,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.01,0.18547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.01,0.24186880588531495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,power_law_1.01,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,power_law_1.01,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,power_law_1.01,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,power_law_1.01,0.032051199674606325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,power_law_1.01,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,power_law_1.01,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,power_law_1.01,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,power_law_1.01,0.1282047986984253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,power_law_1.01,0.18717440366744995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,power_law_1.01,0.22867839336395263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,power_law_1.01,0.3172991991043091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,power_law_1.01,0.39831678867340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,power_law_1.2,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,power_law_1.2,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,power_law_1.2,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,power_law_1.2,0.048851200938224794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,power_law_1.2,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,power_law_1.2,0.05084159970283508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.2,0.08828160166740417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.2,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.2,0.12017279863357544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.2,0.146451199054718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.2,0.20316801071166993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.2,0.2645567893981934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.2,0.4005375862121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.2,0.46640639305114745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,balanced,0.09297066926956177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,balanced,0.13320533434549967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,balanced,0.1369599997997284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,balanced,0.13392000397046408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,balanced,0.1362666686375936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,balanced,0.1360213359196981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,balanced,0.13921067118644714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,balanced,0.1421333352724711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,balanced,0.14220800002415976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,balanced,0.14962666233380637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,balanced,0.15668267011642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,balanced,0.16154133280118307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,balanced,0.17476266622543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,balanced,0.19865065813064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,balanced,0.21493866046269736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,balanced,0.25995733340581256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,balanced,0.29868799448013306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,balanced,0.4228586753209432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,balanced,0.48742401599884033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,balanced,0.7118133703867594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,balanced,0.8779253164927164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,balanced,1.2807146708170574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,balanced,1.6644585927327473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.10789120197296143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.13658879995346068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,balanced,0.09516800443331401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,balanced,0.0995786686738332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,balanced,0.10964266459147136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,balanced,0.13179733355840048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,balanced,0.1830293337504069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,balanced,0.26211732625961304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,balanced,0.3430826663970947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,balanced,0.3506293296813965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,balanced,0.3492746750513713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,balanced,0.35282667477925617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,balanced,0.35316268603007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,balanced,0.3609866698582967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,balanced,0.36236266295115155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,balanced,0.37212268511454266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,balanced,0.3800373474756877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,balanced,0.3879786729812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,balanced,0.4029119809468587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,balanced,0.44948267936706543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,balanced,0.48285865783691406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,balanced,0.562826673189799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,balanced,0.6319573322931925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,balanced,0.7887840270996094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,balanced,0.9413920243581136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,balanced,1.2835040092468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,balanced,1.7589653333028157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.18395520448684693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,balanced,2.431829293568929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,balanced,3.259087880452474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.2306368112564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.3536895990371704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,balanced,0.0912000040213267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,balanced,0.12451199690500896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,balanced,0.1868000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,balanced,0.3174933393796285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,balanced,0.5746719837188721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,balanced,0.7051200071970621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,balanced,0.7125386397043864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,balanced,0.7147839864095052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,balanced,0.7182506720225016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,balanced,0.7209493319193522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,balanced,0.7273386319478353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,balanced,0.7338986396789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,balanced,0.7349279721577963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,balanced,0.7468533515930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.413753604888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,balanced,0.760159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,balanced,0.7718559900919596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,balanced,0.7923626899719238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,balanced,0.833189328511556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.869109312693278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.1307360331217449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.054309368133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,balanced,1.4568799336751301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,balanced,1.4539093971252441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,balanced,2.1413119633992515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,balanced,2.482762654622396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,balanced,3.646944046020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,balanced,4.508426666259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,power_law_1.2,0.0253248006105423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,power_law_1.2,0.023398399353027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,power_law_1.2,0.023609599471092223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,power_law_1.2,0.02744320034980774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,power_law_1.2,0.03033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,power_law_1.2,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,power_law_1.2,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,power_law_1.2,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,power_law_1.2,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,power_law_1.2,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,power_law_1.2,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.11527040004730224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.13829120397567748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.19461120367050172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.23133440017700196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.3054784059524536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.43692798614501954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,balanced,0.05117333432038625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,balanced,0.05538133283456167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,balanced,0.06958933174610138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,balanced,0.07029333213965099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,balanced,0.07102933526039124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,balanced,0.0722453345855077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,balanced,0.07245866457621257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,balanced,0.07247999807198842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,balanced,0.07489066819349925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,balanced,0.07888533174991608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.09905067086219788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.11826133728027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.13779200116793314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.17273066441218057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.21027199427286783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.2733333309491475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.34147198994954425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.46748268604278564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.6043039957682291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,balanced,0.0566293348868688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,balanced,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,balanced,0.1011840005715688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,balanced,0.1216213305791219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,balanced,0.12377599875132243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,balanced,0.12384000420570374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,balanced,0.12546133001645407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,balanced,0.12469333410263062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,balanced,0.1255466639995575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,balanced,0.1277653376261393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,balanced,0.13220266501108804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,balanced,0.1365546683470408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,balanced,0.1376159985860189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,balanced,0.1476586659749349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,balanced,0.16378666957219443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,balanced,0.17485866943995157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,balanced,0.20168532927831015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,balanced,0.22765866915384927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,balanced,0.2834773262341817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,balanced,0.3387306531270345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,balanced,0.4586613178253174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,balanced,0.6142880121866862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,balanced,0.8550399939219157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,balanced,1.1259466807047527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,power_law_1.2,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,power_law_1.2,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,power_law_1.2,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,power_law_1.2,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,power_law_1.2,0.0874559998512268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,power_law_1.2,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,power_law_1.2,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,power_law_1.2,0.09049599766731262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,power_law_1.2,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,power_law_1.2,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,power_law_1.2,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,power_law_1.2,0.11815040111541748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,power_law_1.2,0.12432639598846436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,power_law_1.2,0.13752959966659545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,power_law_1.2,0.16380800008773805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,power_law_1.2,0.16770559549331665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,power_law_1.2,0.2082751989364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,power_law_1.2,0.25360000133514404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,power_law_1.2,0.3384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,power_law_1.2,0.4077951908111572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,power_law_1.2,0.5697663784027099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,power_law_1.2,0.7711679935455322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,power_law_1.2,1.1022080421447753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,power_law_1.2,1.3940735816955567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,power_law_1.01,0.07682560086250305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,power_law_1.01,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,power_law_1.01,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,power_law_1.01,0.11503360271453858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,power_law_1.01,0.13360639810562133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,power_law_1.01,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,power_law_1.01,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,power_law_1.01,0.16647679805755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,power_law_1.01,0.1688256025314331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,power_law_1.01,0.17265280485153198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,power_law_1.01,0.179366397857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,power_law_1.01,0.18065279722213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,power_law_1.01,0.18885120153427123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,power_law_1.01,0.19635839462280275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,power_law_1.01,0.20511360168457032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,power_law_1.01,0.21527040004730225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,power_law_1.01,0.2470911979675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.01,0.27957758903503416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.01,0.3441728115081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.01,0.4079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.01,0.5352255821228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.01,0.6629183769226075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.01,0.9239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.01,1.1813952445983886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.01,1.6981119155883788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.01,2.2159231185913084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,balanced,0.058490668733914696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,balanced,0.03615466753641764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,balanced,0.029146666328112285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,balanced,0.03481066723664602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,balanced,0.03524799893299738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,balanced,0.03700266778469086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,balanced,0.03813866774241129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,balanced,0.08079466720422109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,balanced,0.09062932928403218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.03832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.06952959895133973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.07343999743461609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.0764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08510079979896545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.09199360013008118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.13530880212783813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,power_law_1.01,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.14810880422592163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.18681600093841552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.22070400714874266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.30871679782867434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.3616703987121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,power_law_1.01,0.06335999965667724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.5200511932373046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,power_law_1.01,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.658080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,power_law_1.01,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,power_law_1.01,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,power_law_1.01,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,power_law_1.01,0.08651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,power_law_1.01,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.01,0.10119040012359619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.01,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.01,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.01,0.15927679538726808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.01,0.18526079654693603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.01,0.2579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.01,0.3127680063247681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.01,0.40628480911254883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.01,0.5210239887237549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.16628479957580566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.21763839721679687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.2812096118927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.3963776111602783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,power_law_1.01,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,power_law_1.01,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,power_law_1.01,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,power_law_1.01,0.05997440218925476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,power_law_1.01,0.0606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,power_law_1.01,0.06919040083885193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,power_law_1.01,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,power_law_1.01,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,power_law_1.01,0.09767680168151856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,power_law_1.01,0.11055359840393067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,power_law_1.01,0.13571200370788575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,power_law_1.01,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,power_law_1.01,0.20615038871765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,power_law_1.01,0.2514944076538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,power_law_1.01,0.3313535928726196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,power_law_1.01,0.4116672039031982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,power_law_1.01,0.6429056167602539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,power_law_1.01,0.8735551834106445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.49004158973693845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.08101760149002075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.09350399971008301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.11229439973831176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.13210879564285277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.18786560297012328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.2253567934036255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.28061439990997317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.3771456003189087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.4635327816009521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.7105343818664551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,power_law_1.2,0.8956416130065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,power_law_1.2,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,power_law_1.2,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,power_law_1.2,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,power_law_1.2,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,power_law_1.2,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,power_law_1.2,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,power_law_1.2,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,power_law_1.2,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,power_law_1.2,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,power_law_1.2,0.07705600261688232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.2,0.1168063998222351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.2,0.14236799478530884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.2,0.17461760044097902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.2,0.2560064077377319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,power_law_1.01,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.2,0.2941119909286499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,power_law_1.01,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.2,0.399564790725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,power_law_1.01,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,power_law_1.01,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,power_law_1.01,0.10074880123138427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,power_law_1.01,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,power_law_1.01,0.09824640154838563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,power_law_1.01,0.1096127986907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,power_law_1.01,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,power_law_1.01,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,power_law_1.01,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,power_law_1.01,0.11949440240859985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,power_law_1.01,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,power_law_1.01,0.1392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,power_law_1.01,0.14997119903564454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,power_law_1.01,0.17367039918899535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,power_law_1.01,0.194758403301239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,power_law_1.01,0.23316481113433837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,power_law_1.01,0.27139840126037595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,power_law_1.01,0.3644223928451538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,power_law_1.01,0.43666558265686034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,power_law_1.01,0.6298816204071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.2,0.5832064151763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,power_law_1.01,0.8164416313171386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,balanced,0.03789333254098892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,balanced,0.09731200337409973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,balanced,0.09831999739011128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,balanced,0.10197866956392924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,balanced,0.10331733028093974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,balanced,0.1032426655292511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,balanced,0.10658133029937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,balanced,0.10920533537864685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,balanced,0.11371733744939168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,balanced,0.1147093375523885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,balanced,0.12435733278592427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,balanced,0.13502933581670126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,balanced,0.13865066568056741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,balanced,0.14528533816337585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,balanced,0.17432532707850137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,balanced,0.1946986714998881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,balanced,0.26873600482940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,balanced,0.2945493261019389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,balanced,0.43088531494140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,balanced,0.519599994023641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,balanced,0.761082649230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,balanced,0.9766560395558676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,balanced,1.4427839914957683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,balanced,1.8579786618550618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.01,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.01,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.01,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.01,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.01,0.06737279891967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.01,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.01,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.01,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.01,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.01,0.08970239758491516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.01,0.10386559963226319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.01,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.01,0.1366719961166382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.01,0.15702400207519532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.01,0.2059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.01,0.219596791267395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.01,0.28592638969421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.01,0.3487103939056396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.01,0.4938047885894775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.01,0.588588809967041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,balanced,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,balanced,0.12198932965596516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,balanced,0.1241333285967509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,balanced,0.1258133351802826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,balanced,0.12685333689053854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,balanced,0.12683733304341635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,balanced,0.1335040032863617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,balanced,0.13365866740544638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,balanced,0.1320693294207255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,balanced,0.13587733109792074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,balanced,0.14128533005714417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,balanced,0.1455733378728231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,balanced,0.15761066476504007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,balanced,0.17309866348902384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,balanced,0.18519467115402222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,balanced,0.2153759996096293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,balanced,0.24370666344960532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,balanced,0.32103999455769855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,balanced,0.3681439956029256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,balanced,0.51582932472229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,balanced,0.6300373474756876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,balanced,0.9077866872151693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,balanced,1.1875893274943035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,power_law_1.2,0.07728639841079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,power_law_1.2,0.09211519956588746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,power_law_1.2,0.1023360013961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,power_law_1.2,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,power_law_1.2,0.10759680271148682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,power_law_1.2,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,power_law_1.2,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,power_law_1.2,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,power_law_1.2,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,power_law_1.2,0.10538239479064941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,power_law_1.2,0.1053056001663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,power_law_1.2,0.10391039848327636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,power_law_1.2,0.10775680541992187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,power_law_1.2,0.1132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,power_law_1.2,0.1223039984703064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,power_law_1.2,0.118777596950531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,power_law_1.2,0.12889599800109863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,power_law_1.2,0.14442880153656007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.2,0.156985604763031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.2,0.18744959831237792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.2,0.21799681186676026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.2,0.27099521160125734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.2,0.3422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.2,0.48090238571166993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.2,0.6189695835113526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.2,0.9184767723083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.2,1.3048447608947753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,power_law_1.01,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,power_law_1.01,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,power_law_1.01,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,power_law_1.01,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,power_law_1.01,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,power_law_1.01,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,power_law_1.01,0.0649728000164032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,power_law_1.01,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,power_law_1.01,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,power_law_1.01,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,power_law_1.01,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,power_law_1.01,0.10093439817428589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,power_law_1.01,0.1251520037651062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.01,0.14133119583129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.01,0.1789952039718628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.01,0.20845439434051513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.01,0.30845439434051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.01,0.40464000701904296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.01,0.5121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,balanced,0.045552000403404236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,balanced,0.04719999929269155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,balanced,0.03713600089152654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,balanced,0.03928533444801966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,balanced,0.04127466678619385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,balanced,0.04668800036112467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,balanced,0.06337599953015645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,balanced,0.05762666463851929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,balanced,0.07438399891058604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,balanced,0.0781173308690389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,balanced,0.09659199913342793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,balanced,0.10035199920336406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,balanced,0.13360533118247986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,balanced,0.15205867091814676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,balanced,0.21486934026082358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,balanced,0.25783999760945636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.01,0.6843776226043701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.01,1.1391360282897949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.01,1.2328831672668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,balanced,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,balanced,0.03387733300526937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,balanced,0.03518400092919668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,balanced,0.03596800069014231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,balanced,0.04138133426507314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.06394133468468984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.07021333277225494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.0921493371327718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.11335466305414836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.15018133322397867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.17322667439778647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.16309759616851807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.25627520084381106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.27267839908599856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.28332159519195554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.28984320163726807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.298419189453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3086719989776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.31957120895385743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.3242111921310425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.3291392087936401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.3501312017440796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.35514240264892577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.37625598907470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.42314882278442384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.48743038177490233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.5921984195709229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.6609536170959472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.8334272384643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.058841609954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.3613951683044434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,power_law_1.2,1.7149120330810548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,power_law_1.2,2.268307113647461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,power_law_1.2,2.7794815063476563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,power_law_1.01,0.08072320222854615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,power_law_1.01,0.11721600294113159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,power_law_1.01,0.17411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,power_law_1.01,0.2965759992599487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,power_law_1.01,0.42790398597717283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,power_law_1.01,0.5640704154968261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,power_law_1.01,0.7803071975708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,power_law_1.01,0.7952832221984864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,power_law_1.01,0.8319104194641114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,power_law_1.01,0.857420825958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,power_law_1.01,0.8782591819763184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,power_law_1.01,0.9241087913513184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,power_law_1.01,0.9115967750549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,power_law_1.01,0.9440832138061523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,power_law_1.01,0.9701888084411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,power_law_1.01,0.9778752326965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,power_law_1.01,1.0401920318603515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,power_law_1.01,1.1311871528625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,power_law_1.01,1.102329635620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,power_law_1.01,1.2631168365478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,power_law_1.01,1.255782413482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,power_law_1.01,1.420684814453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,balanced,0.0535093347231547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,power_law_1.01,1.581510353088379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,balanced,0.05451733370621999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,balanced,0.07072533170382182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,balanced,0.07886933286984761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,balanced,0.08255466818809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,balanced,0.08738666772842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,power_law_1.01,1.9388032913208009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,balanced,0.08515733480453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,balanced,0.08716799815495808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,balanced,0.09144533673922221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,balanced,0.0899733304977417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,power_law_1.01,2.1793407440185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,balanced,0.09884267052014668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,balanced,0.10584533214569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,balanced,0.11947733163833618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,balanced,0.1285546620686849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,balanced,0.15387200315793356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,power_law_1.01,2.770591926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,balanced,0.17547200123469034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,balanced,0.23640533288319907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,balanced,0.24761066834131876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,balanced,0.3181866606076558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,balanced,0.3811200062433879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,power_law_1.01,3.388876724243164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,balanced,0.533893346786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,balanced,0.6620853344599406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,balanced,0.04869333406289419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,balanced,0.05393599967161814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,balanced,0.058431997895240784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,balanced,0.06401066482067108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,balanced,0.06803733110427856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,balanced,0.08038933575153351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,balanced,0.08285866677761078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,balanced,0.10967999696731567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,balanced,0.11942399541536967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,balanced,0.1723466714223226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,balanced,0.17116800944010416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,balanced,0.21977599461873373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,balanced,0.2590026656786601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,balanced,0.3484586477279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,balanced,0.42560001214345294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,power_law_1.01,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,power_law_1.01,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,power_law_1.01,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,power_law_1.01,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,power_law_1.01,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,power_law_1.01,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,power_law_1.01,0.115283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,power_law_1.01,0.13803520202636718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,power_law_1.01,0.1815808057785034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,power_law_1.01,0.20375680923461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,power_law_1.01,0.2840768098831177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,power_law_1.01,0.40535678863525393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,balanced,0.042394667863845825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,balanced,0.04409599800904592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,balanced,0.06934933364391327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,balanced,0.09308266639709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,balanced,0.09385599692662557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,balanced,0.09546132882436116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,balanced,0.09823466340700786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,balanced,0.0969599982102712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,balanced,0.09973866740862529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,balanced,0.10220799843470256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,balanced,0.1011253297328949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,balanced,0.10711999734242757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,balanced,0.11198932925860088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,balanced,0.11602666974067688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,balanced,0.12499200304349263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,balanced,0.1426293353239695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,balanced,0.15372266372044882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,balanced,0.18197866280873617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,balanced,0.20482667287190756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,balanced,0.28364266951878864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,balanced,0.32650667428970337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,balanced,0.47447999318440753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,balanced,0.5813653469085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,balanced,0.8574133714040121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,balanced,1.1283306280771892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,2,power_law_1.2,0.15829119682312012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,2,power_law_1.2,0.17978880405426026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,2,power_law_1.2,0.25647358894348143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,2,power_law_1.2,0.3017280101776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,2,power_law_1.2,0.3708928108215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,2,power_law_1.2,0.5768896102905273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,2,power_law_1.2,0.6334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,2,power_law_1.2,0.6221504211425781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,2,power_law_1.2,0.6384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,2,power_law_1.2,0.661836814880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,2,power_law_1.2,0.6793663978576661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,2,power_law_1.2,0.6851967811584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,2,power_law_1.2,0.6751232147216797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,2,power_law_1.2,0.7103487968444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,2,power_law_1.2,0.7105984210968017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,2,power_law_1.2,0.7252543926239013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,2,power_law_1.2,0.7333759784698486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,2,power_law_1.2,0.7739967823028564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.2,0.824185562133789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.2,0.9250752449035644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.2,1.0577664375305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.2,1.3498880386352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.2,1.5727552413940429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.2,2.1290496826171874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.2,2.5476543426513674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.2,3.499923324584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.2,4.58392333984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,power_law_1.2,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,power_law_1.2,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,power_law_1.2,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,power_law_1.2,0.1250175952911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,power_law_1.2,0.1306175947189331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,power_law_1.2,0.13623679876327516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,power_law_1.2,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,power_law_1.2,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,power_law_1.2,0.15111039876937865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,power_law_1.2,0.15232640504837036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,power_law_1.2,0.1583232045173645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,power_law_1.2,0.17144320011138917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,power_law_1.2,0.17891839742660523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,power_law_1.2,0.19288959503173828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,power_law_1.2,0.2241408109664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,power_law_1.2,0.2544640064239502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,power_law_1.2,0.32312960624694825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,power_law_1.2,0.36482560634613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,power_law_1.2,0.48555521965026854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,power_law_1.2,0.5975808143615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.01,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.01,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.01,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.01,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.01,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.01,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.01,0.07356160283088684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.01,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.01,0.09896320104598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.01,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.01,0.19004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.01,0.2066879987716675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.01,0.2566015958786011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.01,0.3205440044403076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.01,0.4289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.01,0.5369088172912597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,power_law_1.2,0.8266688346862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,power_law_1.2,1.0446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,balanced,0.06713066498438518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,balanced,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,balanced,0.13678399721781412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,balanced,0.13959466417630514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,balanced,0.13993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,balanced,0.14081600308418274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,balanced,0.14500266313552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,balanced,0.14383467038472494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,balanced,0.14763200283050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,balanced,0.15067733327547708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,balanced,0.15402133266131082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,balanced,0.1572480003039042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,balanced,0.16339199741681418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,balanced,0.1736639936765035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,balanced,0.19273066520690918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,balanced,0.21647467215855917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,balanced,0.24846933285395303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,balanced,0.2845279971758525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,balanced,0.39618666966756183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,balanced,0.4461546738942464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,balanced,0.668341318766276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,balanced,0.8349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,balanced,1.2113386789957683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,balanced,1.579861323038737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,power_law_1.01,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,power_law_1.01,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,power_law_1.01,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,power_law_1.01,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,power_law_1.01,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,power_law_1.01,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,power_law_1.01,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.01,0.11185920238494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.01,0.13431040048599244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.01,0.17366399765014648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.01,0.21232640743255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.01,0.305132794380188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,power_law_1.2,1.490931224822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.01,0.3807615995407104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,power_law_1.2,1.9743488311767579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.09618560075759888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.1056831955909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.14599679708480834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.18566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.2609407901763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,power_law_1.01,0.30621440410614015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,power_law_1.2,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,power_law_1.2,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,power_law_1.2,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,power_law_1.2,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,power_law_1.2,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,power_law_1.2,0.07115520238876342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,power_law_1.2,0.07330560088157653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,power_law_1.2,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,power_law_1.2,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,power_law_1.2,0.10476800203323364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,power_law_1.2,0.13281279802322388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,power_law_1.2,0.15393919944763185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,power_law_1.2,0.22325758934020995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,power_law_1.2,0.27686400413513185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,power_law_1.2,0.3709503889083862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,power_law_1.2,0.5217919826507569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,power_law_1.2,0.8140543937683106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,power_law_1.2,0.030649599432945252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,power_law_1.2,0.02932479977607727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,power_law_1.2,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,power_law_1.2,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,power_law_1.2,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,power_law_1.2,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,power_law_1.2,0.07080320119857789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,balanced,0.029872000217437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,balanced,0.03618133316437403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,balanced,0.03896533449490865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,balanced,0.03935466706752777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,balanced,0.048698668678601585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,balanced,0.05054399867852529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,balanced,0.08263466755549113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,balanced,0.08950400352478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,balanced,0.1109920044740041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,balanced,0.1341653366883596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,balanced,0.17620267470677695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,balanced,0.21019200483957926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,power_law_1.2,0.09486719965934753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,balanced,0.2880106568336487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,balanced,0.3493013381958008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,power_law_1.2,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,power_law_1.2,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,power_law_1.2,1.1065919876098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,power_law_1.2,0.17957760095596315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,power_law_1.2,0.27000319957733154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,power_law_1.2,0.3150719881057739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,power_law_1.2,0.45863680839538573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,power_law_1.2,0.04871039986610413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,power_law_1.2,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,power_law_1.2,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,power_law_1.2,0.08583040237426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,power_law_1.2,0.10540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,power_law_1.2,0.12010879516601562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,power_law_1.2,0.15242880582809448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.2,0.17607040405273439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.2,0.2537472009658813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.2,0.3007296085357666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,power_law_1.2,0.5096447944641114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.2,0.44318718910217286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.2,0.4984447956085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.2,0.8189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.24276480674743653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.3649535894393921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.46856961250305174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.8381504058837891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9101375579833985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,power_law_1.2,0.9579839706420898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,power_law_1.2,0.9645183563232422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,power_law_1.2,0.9856767654418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.0196096420288085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.0421695709228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.053004837036133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.0888511657714843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.1287872314453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.146457576751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.1848320007324218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.2,1.039072036743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.299507236480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.3688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,power_law_1.2,1.6006784439086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,power_law_1.2,1.6997503280639648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,power_law_1.2,2.0943424224853517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,power_law_1.2,2.1070016860961913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,power_law_1.2,2.55416316986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,power_law_1.2,3.4221630096435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,power_law_1.2,4.333798217773437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.2,1.5284671783447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,power_law_1.2,5.176460647583008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.2,1.914726448059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.11599359512329102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.19749759435653685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.28928000926971437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.3836031913757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.5911104202270507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.6309375762939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6368192195892334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.6490943908691407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.6637695789337158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.6727807998657227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.6877439975738525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.6971072196960449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.7303999900817871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.747814416885376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.7749120235443115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.7531775951385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,power_law_1.01,0.8276991844177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,power_law_1.01,0.8333312034606933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,power_law_1.01,0.9275648117065429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.0218688011169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,power_law_1.01,1.1902079582214355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,power_law_1.01,1.3600383758544923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,power_law_1.01,1.715590476989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,power_law_1.01,2.049203109741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,power_law_1.01,2.762368011474609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,power_law_1.01,3.5219070434570314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.2263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.3192255973815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,power_law_1.01,0.04996480047702789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,power_law_1.01,0.051526397466659546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,power_law_1.01,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,power_law_1.01,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,power_law_1.01,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,power_law_1.01,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,power_law_1.01,0.07599999904632568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,power_law_1.01,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,power_law_1.01,0.09916160106658936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.01,0.12502399682998658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.01,0.14217599630355834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.01,0.17671040296554566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,balanced,0.07237866520881653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.01,0.22069759368896485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.01,0.3064703941345215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.01,0.3982208013534546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.01,0.6054336071014405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,balanced,0.11513066291809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,balanced,0.20036800702412924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,balanced,0.3689546585083008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,balanced,0.37011198202768963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.01,0.7397056102752686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,balanced,0.37276800473531085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,balanced,0.3748053312301636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,balanced,0.3767999807993571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,balanced,0.37749334176381427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,balanced,0.3789706627527873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,balanced,0.3791786829630534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,balanced,0.3834720055262248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,balanced,0.38415467739105225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,balanced,0.39084800084431964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,balanced,0.39764265219370526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,balanced,0.4076426823933919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,balanced,0.4193386634190877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,balanced,0.44734398523966473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,balanced,0.48835734526316327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,balanced,0.49167998631795246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,balanced,0.5872000058492025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,balanced,0.6112106641133627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,balanced,0.8182400067647299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,balanced,0.9065706729888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,balanced,1.2351946830749512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,balanced,1.5891092618306477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.2,0.15852160453796388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.2,0.24243199825286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.2,0.3188479900360107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.2,0.41751680374145506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.2,0.6679423809051513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.2,0.7082111835479736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.2,0.6909696102142334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.2,0.725548791885376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.2,0.7390912055969239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.2,0.7728831768035889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.2,0.7898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.2,0.7941952228546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.2,0.8321663856506347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.2,0.8681407928466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.2,0.8996224403381348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.2,0.9284031867980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.2,1.0920767784118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.2,1.0315584182739257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.2,1.201369571685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.2,1.367155170440674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.2,1.6038528442382813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.2,1.651033592224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.06113280057907104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.2,2.027769660949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.2,2.531328010559082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.2,3.511219024658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.16739200353622435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.19364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.2724607944488525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.3333888053894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.5081600189208985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.6269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.2,4.031411361694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,balanced,0.0440586656332016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,balanced,0.045519997676213585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,balanced,0.06214933097362518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,balanced,0.06649599969387054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,balanced,0.06587199866771698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,balanced,0.07232533395290375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,balanced,0.07693866888682048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,balanced,0.08451732993125916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,balanced,0.09041600426038106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,balanced,0.10314666231473286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,balanced,0.11980799833933513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,balanced,0.15324800213178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,balanced,0.18082666397094727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,balanced,0.24327999353408813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,balanced,0.28784533341725665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,balanced,0.40330131848653156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,balanced,0.5129760106404623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,power_law_1.01,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,power_law_1.01,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,power_law_1.01,0.1476863980293274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,power_law_1.01,0.22167680263519288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,power_law_1.01,0.2921472072601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,power_law_1.01,0.45822720527648925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,power_law_1.01,0.4887807846069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,power_law_1.01,0.4960063934326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,power_law_1.01,0.5226111888885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,power_law_1.01,0.5265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,power_law_1.01,0.5435647964477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,power_law_1.01,0.5490623950958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,power_law_1.01,0.5618495941162109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,power_law_1.01,0.5872960090637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,power_law_1.01,0.6172736167907715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,power_law_1.01,0.6194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,power_law_1.01,0.6558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,power_law_1.01,0.7231808185577393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,power_law_1.01,0.7620031833648682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,power_law_1.01,0.8919936180114746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,power_law_1.01,0.998969554901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.11963520050048829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,power_law_1.01,1.249721622467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.14506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,power_law_1.01,1.3576576232910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.21015679836273193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,power_law_1.01,1.7721920013427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.29052159786224363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.38368639945983884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,power_law_1.01,2.4708480834960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.4365248203277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.5833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,power_law_1.01,3.248767852783203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.5994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.6090879917144776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,power_law_1.01,3.988723373413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.6467328071594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.6776000022888183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.6839680194854736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.7190144062042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.7512896060943604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.7682176113128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.8191103935241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.8916223526000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,0.9541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.1133055686950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.206777572631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,1.445241641998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,1.7344831466674804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,2.1877248764038084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.2,2.607807922363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.2,3.4615745544433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.2,4.281881713867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,balanced,0.03382399926582972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,balanced,0.03401600072781245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,balanced,0.035904000202814736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,balanced,0.04808000226815542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,balanced,0.052298665046691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,balanced,0.07014933228492737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,balanced,0.10646399855613708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,balanced,0.09331732988357544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,balanced,0.09643733501434326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,balanced,0.09514666597048442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,balanced,0.09494933485984802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,balanced,0.09660800298055013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,balanced,0.09781333804130554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,balanced,0.09977066516876221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,balanced,0.10274133086204529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,balanced,0.10315199693044026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,balanced,0.10875733693440755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,balanced,0.1111893355846405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,balanced,0.11769599715868632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,balanced,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.13661866386731467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.17062934239705405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.21961599588394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.2519200046857198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.34122665723164874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.4100266695022583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.5766666730244955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.7410879929860433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,power_law_1.2,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,power_law_1.2,0.11119359731674194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,power_law_1.2,0.14288640022277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,power_law_1.2,0.16078720092773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,power_law_1.2,0.25698559284210204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,power_law_1.2,0.2848063945770264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,power_law_1.2,0.2785151958465576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,power_law_1.2,0.2928639888763428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,power_law_1.2,0.30145280361175536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,power_law_1.2,0.2953984022140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,power_law_1.2,0.3075455904006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,power_law_1.2,0.31626238822937014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,power_law_1.2,0.33292160034179685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,power_law_1.2,0.33816959857940676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,power_law_1.2,0.3582848072052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,power_law_1.2,0.35687038898468015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,power_law_1.2,0.40204801559448244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,power_law_1.2,0.4442495822906494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,power_law_1.2,0.5253183841705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,power_law_1.2,0.589792013168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,power_law_1.2,0.747980785369873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,power_law_1.2,0.8028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,power_law_1.2,1.0451647758483886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,power_law_1.2,1.3602496147155763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,power_law_1.01,0.15555839538574218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,power_law_1.01,0.2346112012863159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,power_law_1.01,0.23111040592193605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,power_law_1.01,0.23434879779815673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,power_law_1.01,0.24787840843200684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,power_law_1.01,0.25017600059509276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,power_law_1.01,0.2545919895172119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,power_law_1.01,0.25462400913238525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,power_law_1.01,0.25040640830993655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,power_law_1.01,0.2683072090148926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,power_law_1.01,0.26825599670410155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,power_law_1.01,0.2874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,power_law_1.01,0.32678399085998533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,power_law_1.01,0.3348671913146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,power_law_1.01,0.33016319274902345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,power_law_1.01,0.3846015930175781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,power_law_1.01,0.44490880966186525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,power_law_1.01,0.5526463985443115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,power_law_1.01,0.6513855934143067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,power_law_1.01,0.8635199546813965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,power_law_1.01,1.0829759597778321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,power_law_1.2,1.9515775680541991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,power_law_1.01,1.507033634185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,power_law_1.2,2.530175971984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,power_law_1.01,1.9434879302978516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,power_law_1.01,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,power_law_1.01,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,power_law_1.01,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,power_law_1.01,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,power_law_1.01,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,power_law_1.01,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,power_law_1.01,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,power_law_1.01,0.11236480474472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,power_law_1.01,0.11908479928970336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,power_law_1.01,0.1462399959564209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,power_law_1.01,2.8350719451904296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,power_law_1.01,0.17023359537124633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,balanced,0.05878399809201559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,balanced,0.05844266712665558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,balanced,0.0569653312365214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,balanced,0.08326399823029836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,balanced,0.09047466516494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,balanced,0.09160533547401428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,balanced,0.09488000472386678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,balanced,0.09498666723569234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,balanced,0.09483733773231506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,balanced,0.09604799747467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,balanced,0.09690133730570476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,balanced,0.1042080024878184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,balanced,0.10700800021489461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,balanced,0.10914666453997295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,balanced,0.11291199922561646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,balanced,0.1183519959449768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,balanced,0.13102933764457703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,balanced,0.1405173341433207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,balanced,0.1872053345044454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,balanced,0.20459733406702676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,balanced,0.2651360034942627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,balanced,0.3332586685816447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,balanced,0.4272426764170329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,power_law_1.01,0.23349759578704835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,balanced,0.5639679829279581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,power_law_1.01,0.28255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,power_law_1.01,0.40289921760559083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,power_law_1.01,3.7166080474853516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,power_law_1.01,0.5106688022613526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,power_law_1.01,0.7328959941864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,power_law_1.01,0.9498944282531738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,power_law_1.2,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,power_law_1.2,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,power_law_1.2,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,power_law_1.2,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,power_law_1.2,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,power_law_1.2,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.2,0.11210240125656128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.2,0.1360703945159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.2,0.18216960430145263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.2,0.24680960178375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,balanced,0.03439466655254364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,balanced,0.03465600063403448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,balanced,0.03382399926582972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,balanced,0.03595199932654699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,balanced,0.04668800036112467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,balanced,0.05365866422653198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,balanced,0.07077333331108093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,balanced,0.09310400485992432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,balanced,0.12959999839464822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,balanced,0.15452266732851663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,balanced,0.20478399594624838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,balanced,0.26286399364471436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,balanced,0.04858666658401489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,balanced,0.052416001756985985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.2,0.3222527980804443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,balanced,0.09629866480827332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,balanced,0.13110400239626566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,balanced,0.16332266728083292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,balanced,0.16359466314315796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,balanced,0.16311466693878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,balanced,0.16475199659665427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,balanced,0.16639999548594156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,balanced,0.16804800430933634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,balanced,0.1710453430811564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,balanced,0.17200533548990884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,balanced,0.1774239937464396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,balanced,0.18372799952824911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,balanced,0.18699200948079428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,balanced,0.19549334049224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,balanced,0.21957866350809732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,balanced,0.22950400908788046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,balanced,0.29758934179941815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,balanced,0.3179840048154195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,balanced,0.46323732535044354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,balanced,0.4718506733576457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,balanced,0.676736036936442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,balanced,0.7803093592325846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,balanced,1.1335840225219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,balanced,1.3889813423156738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.2,0.4966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.2,0.7162496089935303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.2,0.9870976448059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.2,1.3202688217163085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,balanced,0.04207466542720795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,balanced,0.04858666658401489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,balanced,0.06411733229955037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,balanced,0.09074667096138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,balanced,0.11953600247701009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,balanced,0.151119997104009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,balanced,0.21237866083780924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,balanced,0.2577333251635234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,balanced,0.03798400113979975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,balanced,0.04002666721741358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,balanced,0.040261333187421165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.08617066343625386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.09783466657002766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.12454932928085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.1397813359896342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.18920532862345377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.2360159953435262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.33669865131378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.4330879847208659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.03153280019760132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,balanced,0.029338667790095013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,balanced,0.0769706666469574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,balanced,0.07602666815121968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,balanced,0.07708266874154408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,balanced,0.07670400043328603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,balanced,0.07629333436489105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,balanced,0.0780213326215744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,balanced,0.08066133161385854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,balanced,0.08321066697438557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,balanced,0.08559999863306682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.0765504002571106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.09122560024261475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.11370879411697388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.1503424048423767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.19725439548492432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,balanced,0.08850666880607605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,balanced,0.08989333113034566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,balanced,0.0946560005346934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,balanced,0.10689066847165425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,balanced,0.11712533235549927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,balanced,0.13904000322024027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,balanced,0.1567200024922689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,balanced,0.20843199888865152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,balanced,0.23404266436894736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,balanced,0.31854933500289917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,balanced,0.40194666385650635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.2,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.2,0.1119871973991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.2,0.11287039518356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.2,0.115065598487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.2,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.2,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.2,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.2,0.13071999549865723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.2,0.13636480569839476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.2,0.15304960012435914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.2,0.17763839960098265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.2,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.2,0.19089920520782472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.2,0.21258881092071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.2,0.25693440437316895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.2,0.2548288106918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.2,0.33120639324188234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.2,0.4153791904449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.2,0.5487616062164307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.2,0.627123212814331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,power_law_1.01,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,power_law_1.01,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,power_law_1.01,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,power_law_1.01,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,power_law_1.01,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,power_law_1.01,0.05072000026702881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,power_law_1.01,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,power_law_1.01,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.01,0.08767359852790832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.01,0.0979968011379242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.01,0.1281280040740967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.01,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.01,0.20722560882568358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.01,0.2744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.01,0.3728447914123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.01,0.5051648139953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,balanced,0.04456000030040741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,balanced,0.07357333103815715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,balanced,0.08418132861455281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,balanced,0.0888320008913676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,balanced,0.11222933729489644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,balanced,0.1299253304799398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,balanced,0.17062399784723917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,balanced,0.21252266565958658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,balanced,0.2946880062421163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,balanced,0.3728426694869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,power_law_1.2,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,power_law_1.2,0.07002879977226258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,power_law_1.2,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,power_law_1.2,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,power_law_1.2,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,power_law_1.2,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,power_law_1.2,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,power_law_1.2,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,power_law_1.2,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,power_law_1.2,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,power_law_1.2,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,power_law_1.2,0.11473280191421509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.2,0.12247040271759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.2,0.14781440496444703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.2,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.2,0.2134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.2,0.31775360107421874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.2,0.41246719360351564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.2,0.5434624195098877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.2,0.871116828918457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.2,1.203110408782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.10106879472732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.10182399749755859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.11206400394439697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.11538560390472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.11758719682693482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.1295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.14082560539245606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.1454208016395569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.1575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.17777279615402222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.20832641124725343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,balanced,0.029717333614826202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,balanced,0.03385599950949351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,balanced,0.035989334185918175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,balanced,0.04312533140182495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,balanced,0.04001066585381826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.2756927967071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,balanced,0.0664213349421819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,balanced,0.09045867125193278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,balanced,0.11529599626859029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,balanced,0.13989333311716715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,balanced,0.18742932875951132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,balanced,0.2413333257039388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.2827647924423218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.43645439147949217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.5903232097625732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.7560704231262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.0580544471740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,1.6654783248901368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.09784319996833801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.0966592013835907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.1041856050491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.10699520111083985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.12526079416275024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.12750719785690307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.14347519874572753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.19160319566726686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.22232959270477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.28444159030914307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.35038080215454104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.4862847805023193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.6036543846130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.8490688323974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,1.140659236907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,balanced,0.03409600009520849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,balanced,0.03712533414363861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,balanced,0.04021333406368891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,balanced,0.04223999877770742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,balanced,0.04121066629886627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,balanced,0.05583466589450836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,balanced,0.07603199779987335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,balanced,0.08041066428025563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,balanced,0.09528533617655437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,balanced,0.11081066727638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,balanced,0.14290666580200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,balanced,0.17128000656763712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,balanced,0.23431466023127237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,balanced,0.28841066360473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,balanced,0.4118560155232747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,balanced,0.5263520081837972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,1.8785472869873048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,power_law_1.2,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,power_law_1.2,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,power_law_1.2,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,power_law_1.2,0.09996799826622009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,power_law_1.2,0.13131519556045532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,power_law_1.2,0.12394239902496337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,power_law_1.2,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,power_law_1.2,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,power_law_1.2,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,power_law_1.2,0.1409152030944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,power_law_1.2,0.1549888014793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,power_law_1.2,0.16295039653778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,power_law_1.2,0.17366399765014648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,power_law_1.2,0.21120638847351075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,power_law_1.2,0.21224958896636964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,power_law_1.2,0.2683135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,power_law_1.2,0.306662392616272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,power_law_1.2,0.3912384033203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,power_law_1.2,0.5110208034515381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,power_law_1.2,0.7032896041870117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,power_law_1.2,0.8398336410522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,power_law_1.2,1.212275218963623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,power_law_1.2,1.567103958129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.1713215947151184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,power_law_1.2,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.21790719032287598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,power_law_1.2,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.3259007930755615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,power_law_1.2,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.40799360275268554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.5711232185363769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,power_law_1.2,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,power_law_1.2,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,power_law_1.2,0.9138751983642578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,power_law_1.2,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,power_law_1.2,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,power_law_1.2,0.9755647659301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,power_law_1.2,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,power_law_1.2,0.943455982208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,power_law_1.2,0.07246080040931702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,power_law_1.2,0.9935872077941894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,power_law_1.2,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,power_law_1.2,1.0310144424438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,power_law_1.2,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,power_law_1.2,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.0284159660339356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,power_law_1.2,0.1364416003227234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,power_law_1.2,0.1552832007408142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.0432640075683595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.0843647956848144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.1024064064025878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,power_law_1.2,0.20040318965911866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.1716671943664552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.1728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.2290944099426269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.3452223777770995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,power_law_1.2,1.4463168144226075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,power_law_1.2,0.2454848051071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,power_law_1.2,1.6759679794311524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,power_law_1.2,1.7832128524780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,power_law_1.2,2.090176010131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,power_law_1.2,2.1576128005981445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,power_law_1.2,2.6264448165893555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,power_law_1.2,0.36430718898773196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,power_law_1.2,0.4857344150543213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.01,0.19214080572128295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.01,0.2453696012496948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.01,0.24055039882659912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,power_law_1.2,3.3415809631347657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.01,0.24583680629730226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.01,0.24603519439697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.01,0.240883207321167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.01,0.22478721141815186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.01,0.23240959644317627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.01,0.23215360641479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.01,0.23157761096954346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.01,0.22856318950653076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.01,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.01,0.23284480571746827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.01,0.24556798934936525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.01,0.2576512098312378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.01,0.2594496011734009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.01,0.27808640003204343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.01,0.31249279975891114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.01,0.3216768026351929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.01,0.3958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.01,0.44614400863647463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.01,0.5467072010040284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.01,0.5304128170013428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.01,0.7048448085784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.01,0.9134783744812012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,power_law_1.2,4.277273559570313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.01,1.2148736000061036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.01,1.382688045501709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,power_law_1.2,5.104774475097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,power_law_1.2,0.7008831977844239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,power_law_1.2,0.9537216186523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.11073919534683227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.11235840320587158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.11912319660186768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.12339199781417846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.1252992033958435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.12634880542755128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.12982399463653566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.14714879989624025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.1481152057647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.15416959524154664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.17404799461364745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.17084800004959105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.20152320861816406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.23999359607696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.2945791959762573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.33152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.4382656097412109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.2,0.5271423816680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.2,0.7339072227478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.2,0.9408191680908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,power_law_1.01,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,power_law_1.01,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,power_law_1.01,0.029452800750732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,power_law_1.01,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,power_law_1.01,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,power_law_1.01,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,power_law_1.01,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,power_law_1.01,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,power_law_1.01,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,power_law_1.01,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.01,0.1066431999206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.01,0.13383040428161622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.01,0.17148159742355346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.01,0.22683520317077638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.01,0.29838080406188966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.01,0.4308159828186035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.01,0.565715217590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.01,0.9936639785766601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,balanced,0.05134933193524679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,balanced,0.0897759993871053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,balanced,0.0920853316783905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,balanced,0.09406399726867676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,balanced,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,balanced,0.0976586639881134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,balanced,0.0997173289457957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,balanced,0.10180800159772237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,balanced,0.10707733035087585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,balanced,0.1156160036722819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,balanced,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,balanced,0.1265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,balanced,0.13942399621009827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,balanced,0.1502240002155304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,balanced,0.1783626675605774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,balanced,0.20498132705688477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,balanced,0.27777065833409625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,balanced,0.3253706693649292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,balanced,0.4678613344828288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,balanced,0.5800853172938029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,balanced,0.8509493668874105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,balanced,1.1163626511891682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.01,1.357087993621826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,power_law_1.2,0.06936960220336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,power_law_1.2,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,power_law_1.2,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,power_law_1.2,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,power_law_1.2,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,power_law_1.2,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,power_law_1.2,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,power_law_1.2,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,power_law_1.2,0.10819200277328492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.2,0.11978880167007447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.2,0.13953919410705568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.2,0.16328320503234864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.2,0.22181119918823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.2,0.2724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.2,0.40341758728027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.2,0.5072959899902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.2,0.7723392009735107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.2,1.0572223663330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.11719039678573609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.12318719625473022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.12698240280151368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.12894079685211182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.1328384041786194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.13961600065231322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.1436095952987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.15574400424957274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.18280320167541503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.20130560398101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.251859188079834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.2981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.4040063858032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.5076799869537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.7082623958587646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.9067135810852051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.305305576324463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.6953920364379882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,power_law_1.01,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,power_law_1.01,0.025574401021003723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,power_law_1.01,0.02483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,power_law_1.01,0.027039998769760133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,power_law_1.01,0.027500799298286437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,power_law_1.01,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,power_law_1.01,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.10557440519332886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.1282495975494385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.1695680022239685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.2535680055618286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.3151423931121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.3809407949447632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,power_law_1.01,0.0695743978023529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,power_law_1.01,0.11757440567016601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,power_law_1.01,0.11659519672393799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,power_law_1.01,0.11788159608840942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,power_law_1.01,0.12499200105667115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,power_law_1.01,0.12663040161132813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,power_law_1.01,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,power_law_1.01,0.12751359939575196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,power_law_1.01,0.13393919467926024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,power_law_1.01,0.13608319759368898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,power_law_1.01,0.14149760007858275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,power_law_1.01,0.15626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,power_law_1.01,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,power_law_1.01,0.16869759559631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,power_law_1.01,0.1946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,power_law_1.01,0.2147007942199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,power_law_1.01,0.27482240200042723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,power_law_1.01,0.3237760066986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,power_law_1.01,0.4176767826080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,power_law_1.01,0.516377592086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,power_law_1.01,0.7183167934417725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,power_law_1.01,0.9187135696411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,balanced,0.03323200096686681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,balanced,0.03597866743803024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,balanced,0.044405331214269005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,balanced,0.06709333260854085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,balanced,0.07041599849859874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,balanced,0.07156266768773396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,balanced,0.07141333321730296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,balanced,0.07308266560236613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,balanced,0.07492800056934357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,balanced,0.07623999814192454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,balanced,0.0815413345893224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,balanced,0.09047466516494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,balanced,0.09226133426030476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,balanced,0.09584533174832661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,balanced,0.11142399907112122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,balanced,0.12306132912635803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,balanced,0.1588266690572103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,balanced,0.17428799470265707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,balanced,0.24638400475184122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,balanced,0.3036853273709615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,balanced,0.4370986620585124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,balanced,0.5703253348668417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,balanced,0.8294293085734049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,balanced,1.0422826608022053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,power_law_1.01,1.3208127975463868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,power_law_1.01,1.6931135177612304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,power_law_1.2,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,power_law_1.2,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,power_law_1.2,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,power_law_1.2,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,power_law_1.2,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,power_law_1.2,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,power_law_1.2,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.2,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.2,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.2,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.2,0.16384639739990234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.2,0.22086400985717775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.2,0.29322240352630613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.2,0.41095681190490724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.2,0.5754816055297851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.2,0.8364095687866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,balanced,0.05124266445636749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,balanced,0.052815998593966164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,balanced,0.061247999469439186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,balanced,0.062421331803003945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,balanced,0.06612266600131989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,balanced,0.06406933565934499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,balanced,0.06674133241176605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,balanced,0.07097599903742473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,balanced,0.08646399776140849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,balanced,0.08685333530108134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,balanced,0.08621866504351298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,balanced,0.09687466422716777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,balanced,0.10652266939481099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,balanced,0.12575466434160867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,balanced,0.1405333379904429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,balanced,0.1808799902598063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,balanced,0.20736000935236612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,balanced,0.26525332530339557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,balanced,0.3259039918581645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,balanced,0.4416266679763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,balanced,0.5609013239542643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.06307839751243591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.12488319873809814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.13407360315322875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.13759360313415528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.13048959970474244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.14008320569992067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.14666880369186402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.1525823950767517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.15186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.1678272008895874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.18290560245513915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.19687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.23953919410705565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.27070720195770265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.34577920436859133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.41480321884155275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.5666175842285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.6877823829650879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,0.9990976333618165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.4390463829040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,balanced,0.07037333150704701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,balanced,0.14754133423169455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,balanced,0.2563413381576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,balanced,0.4697120189666748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,balanced,0.5730453332265218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,balanced,0.57532799243927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,balanced,0.578170657157898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,balanced,0.58133331934611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,balanced,0.581775983174642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,balanced,0.5877866744995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,balanced,0.5947200059890747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,balanced,0.592469334602356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,balanced,0.600389321645101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,balanced,0.6116106510162354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,balanced,0.6147679885228475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,balanced,0.6289439996083578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,balanced,0.723082701365153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,balanced,0.6729333400726318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,balanced,0.8089386622111002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,balanced,0.7761173248291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,balanced,1.1528159777323406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,balanced,0.9903733730316162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,balanced,1.5044639905293782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,balanced,1.5304586092631023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,balanced,2.514021396636963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,balanced,2.777344067891439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.06113280057907104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.0679423987865448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.07472000122070313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.1150272011756897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.12929279804229737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.13709440231323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.1593791961669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.19606399536132812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.25701758861541746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.3631232023239136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.45919361114501955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.635532808303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.8365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.06464639902114869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.08621439933776856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.12936960458755492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.15324800014495848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.19573119878768921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.20995841026306153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.2092992067337036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.21347200870513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.21674880981445313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.21861119270324708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.22899200916290283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.2362816095352173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.2493824005126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.25268480777740476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.27025279998779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.3033983945846558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.3425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.41786880493164064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.4427199840545654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.5619264125823975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.6729919910430908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,0.8744640350341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.1032704353332519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,1.6013887405395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,2.0294143676757814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.0083776473999024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,balanced,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,balanced,0.029637334247430164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,balanced,0.03494933247566223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,balanced,0.036090667049090065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,balanced,0.03510933369398117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,balanced,0.04462933540344238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,balanced,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,balanced,0.07923733194669087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,balanced,0.1157919963200887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,balanced,0.14408533771832785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,balanced,0.19010132551193237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,balanced,0.22664533058802286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,balanced,0.3226026693979899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,balanced,0.4299573500951131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.15548160076141357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.19795199632644653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.26952319145202636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.4086912155151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.5569024085998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.8758527755737304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.9324992179870606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.9679871559143066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.9871104240417481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,1.0114879608154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,1.031923198699951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,1.0584128379821778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,1.080019187927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,1.1212287902832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,1.1529600143432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,1.1882816314697267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,1.2500736236572265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.3825984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.5093567848205567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,1.8235328674316407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,1.966771125793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,2.48220157623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,2.9791423797607424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,3.980352020263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,5.025888061523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.01,7.198143768310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.01,8.959449768066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,power_law_1.01,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,power_law_1.01,0.07932159900665284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,power_law_1.01,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,power_law_1.01,0.09637759923934937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,power_law_1.01,0.09239680171012879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,power_law_1.01,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,balanced,0.04018666595220566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,balanced,0.04664533336957296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,balanced,0.04545066754023234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,balanced,0.04643733302752177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,balanced,0.06625600159168243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,balanced,0.0721973329782486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,balanced,0.11156266927719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,balanced,0.1290773351987203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,balanced,0.16525866587956747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,balanced,0.20026665925979614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,balanced,0.2829280098279317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,balanced,0.354751984278361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,power_law_1.01,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,power_law_1.01,0.0976960003376007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,power_law_1.01,0.09401599764823913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,power_law_1.01,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,power_law_1.01,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,power_law_1.01,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,power_law_1.01,0.1155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,power_law_1.01,0.12577279806137084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,power_law_1.01,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,power_law_1.01,0.1515712022781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,power_law_1.01,0.18645119667053223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,power_law_1.01,0.21412479877471924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,power_law_1.01,0.2762176036834717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,power_law_1.01,0.3476927995681763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,power_law_1.01,0.46771841049194335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,power_law_1.01,0.6081727981567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,power_law_1.01,0.8415360450744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,power_law_1.01,1.0826751708984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,power_law_1.2,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,power_law_1.2,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,power_law_1.2,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,power_law_1.2,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,power_law_1.2,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,power_law_1.2,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,power_law_1.2,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,power_law_1.2,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,power_law_1.2,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,power_law_1.2,0.10689280033111573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,power_law_1.2,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,power_law_1.2,0.15128320455551147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,power_law_1.2,0.1872704029083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,power_law_1.2,0.232588791847229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,power_law_1.2,0.3015872001647949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,power_law_1.2,0.4226687908172607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,power_law_1.2,0.5615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,balanced,0.029391999046007793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,balanced,0.029472000896930695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,balanced,0.04289066791534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,balanced,0.055493334929148354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,balanced,0.06986133257548015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,balanced,0.08956799904505412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,balanced,0.11427733302116394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,balanced,0.1506613294283549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,balanced,0.18778133392333984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,balanced,0.2629706660906474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,balanced,0.3250880042711894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,balanced,0.4739893277486165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,balanced,0.6105066537857056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,balanced,0.04821333289146423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,balanced,0.05136533578236898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,balanced,0.0529120018084844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,balanced,0.05426133175690969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,balanced,0.054570664962132774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,balanced,0.05730666716893514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,balanced,0.0621973325808843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,balanced,0.06862399975458781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,balanced,0.07227733234564464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,balanced,0.08065066734949748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,balanced,0.08871466914812724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,balanced,0.1030453344186147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,balanced,0.11946133772532146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,balanced,0.15824533502260843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,balanced,0.19271999597549438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,balanced,0.26259199778238934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,balanced,0.3304373423258464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,power_law_1.01,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,power_law_1.01,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,power_law_1.01,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,power_law_1.01,0.06980479955673217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,power_law_1.01,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,power_law_1.01,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,power_law_1.01,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,power_law_1.01,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,power_law_1.01,0.13952640295028687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,power_law_1.01,0.16461440324783325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,power_law_1.01,0.21847679615020751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,power_law_1.01,0.26151039600372317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,power_law_1.01,0.36252799034118655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,power_law_1.01,0.4617472171783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.09625599980354309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.11132160425186158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.1476415991783142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.1896064043045044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,power_law_1.01,0.6536640167236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.21440000534057618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.27457280158996583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,power_law_1.01,0.8471296310424805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.40215678215026857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,balanced,0.08077333370844524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,balanced,0.10605333248774211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,balanced,0.10459199547767639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,balanced,0.10604266325632732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,balanced,0.10718400279680888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,balanced,0.10733333230018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,balanced,0.11105066537857056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,balanced,0.11101866761843364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,balanced,0.1174773375193278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,balanced,0.11702932914098103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,balanced,0.11598933736483256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,balanced,0.11918399731318156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,balanced,0.1283626655737559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,balanced,0.13453867038091025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,balanced,0.15383999546368918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,balanced,0.16878400246302286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,balanced,0.2289066712061564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,balanced,0.26866666475931805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,balanced,0.38277331988016766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,balanced,0.49695467948913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,balanced,0.717583974202474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,balanced,0.9426986376444498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,power_law_1.01,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,power_law_1.01,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,power_law_1.01,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,power_law_1.01,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,power_law_1.01,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,power_law_1.01,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,power_law_1.01,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,power_law_1.01,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,power_law_1.01,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,power_law_1.01,0.09542400240898133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,power_law_1.01,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,power_law_1.01,0.11391999721527099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.01,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.01,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.5455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.01,0.1639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.01,0.22099199295043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.01,0.2608256101608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.01,0.36981759071350095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.01,0.4663871765136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.01,0.6959936141967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.01,0.8981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.6887872219085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,power_law_1.01,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,power_law_1.01,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,power_law_1.01,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,power_law_1.01,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,power_law_1.01,0.07354879975318909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,power_law_1.01,0.07686399817466735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,power_law_1.01,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,power_law_1.01,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,power_law_1.01,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,power_law_1.01,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,power_law_1.01,0.12001919746398926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,power_law_1.01,0.14558080434799195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,power_law_1.01,0.18069119453430177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,power_law_1.01,0.23189759254455566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,power_law_1.01,0.3449984073638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,power_law_1.01,0.43352317810058594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,power_law_1.01,0.6406527996063233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,power_law_1.01,0.8346559524536132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,power_law_1.2,0.020761600136756896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,power_law_1.2,0.02160000056028366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,power_law_1.2,0.021657599508762358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,power_law_1.2,0.023180800676345825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,power_law_1.2,0.023455999791622162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,power_law_1.2,0.024352000653743745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,power_law_1.2,0.024883200228214265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,power_law_1.2,0.02754560112953186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,power_law_1.2,0.027110400795936584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,power_law_1.2,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,power_law_1.2,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,power_law_1.2,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,power_law_1.2,0.029145601391792297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,power_law_1.2,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.1431872010231018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.16735999584197997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,power_law_1.2,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,power_law_1.2,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,power_law_1.2,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,power_law_1.2,0.08244479894638061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,power_law_1.2,0.09582719802856446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,power_law_1.2,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,power_law_1.2,0.10051840543746948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,power_law_1.2,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,power_law_1.2,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.1941823959350586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,power_law_1.2,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,power_law_1.2,0.10650240182876587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,power_law_1.2,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,power_law_1.2,0.11697920560836791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,power_law_1.2,0.1176192045211792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,power_law_1.2,0.1261247992515564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,power_law_1.2,0.14292479753494264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.2,0.16295679807662963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.2,0.18691840171813964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.2,0.23416318893432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.2,0.308569598197937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.2,0.4183551788330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.10677119493484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.11430399417877198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.11818239688873292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.12184959650039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.12414720058441162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.13211519718170167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.13207039833068848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.16194560527801513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.1736448049545288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.21176319122314452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.2151103973388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.2,0.6164735794067383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.29509758949279785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.31578240394592283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,power_law_1.2,0.4699456214904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,power_law_1.2,0.47409920692443847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,power_law_1.2,0.8040512084960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.2,0.8555328369140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.2,1.1538496017456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,power_law_1.2,1.0130623817443847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.2,1.4505215644836427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.06343039870262146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.08111360073089599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.08977919816970825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.09648000001907349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.11334400177001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.1587648034095764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.20018560886383058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.28024959564208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,power_law_1.2,0.3815167903900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,power_law_1.2,0.4806848049163818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,power_law_1.01,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,power_law_1.01,0.0573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,power_law_1.01,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,power_law_1.01,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,power_law_1.01,0.09151359796524047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,power_law_1.01,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,power_law_1.01,0.1056831955909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,power_law_1.01,0.10852479934692383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,power_law_1.01,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,power_law_1.01,0.110534405708313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,power_law_1.01,0.11160320043563843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,power_law_1.01,0.12112640142440796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,power_law_1.01,0.12970240116119386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,power_law_1.01,0.14931199550628663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,power_law_1.01,0.1761728048324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,power_law_1.01,0.17576320171356202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,power_law_1.01,0.2197632074356079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,power_law_1.01,0.24929280281066896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,power_law_1.01,0.33546879291534426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,power_law_1.01,0.41004161834716796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,power_law_1.01,0.5793407917022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,power_law_1.01,0.7094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,power_law_1.01,1.010912036895752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,power_law_1.01,1.3333312034606934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,power_law_1.2,0.6773248195648194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.08199679851531982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.09579520225524903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.09336959719657897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.0943231999874115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.10254720449447632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.10055040121078491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.10027519464492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.10969599485397338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.1198591947555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.13502720594406128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.1541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.1763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.20405120849609376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.24583039283752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.3026047945022583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.4168064117431641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,power_law_1.2,0.4888576030731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,power_law_1.2,0.74235520362854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,power_law_1.2,0.8455167770385742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.09603840112686157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.12533119916915894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.18755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.27746560573577883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.3508991956710815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5739903926849366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.5955520153045655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.609932804107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6393856048583985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6765888214111329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.6675136089324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.690393590927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.6976448059082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.738924789428711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.77358078956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.8428159713745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9155136108398437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.0756544113159179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.176153564453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.4324799537658692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.630668830871582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,2.1255615234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,2.7545024871826174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,3.741766357421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,4.681439971923828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,1,balanced,0.14296000202496847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,1,balanced,0.14841600259145102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,1,balanced,0.1768959959348043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,1,balanced,0.23704532782236734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,1,balanced,0.37299732367197674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,1,balanced,0.4378986756006877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,1,balanced,0.4410133361816406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,1,balanced,0.44653332233428955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,1,balanced,0.44491732120513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,1,balanced,0.44812798500061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,1,balanced,0.4556159973144531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,1,balanced,0.46059199174245197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,1,balanced,0.4671573241551717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,1,balanced,0.4772160053253174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,1,balanced,0.47813868522644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,1,balanced,0.48727468649546307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,1,balanced,0.5057226816813151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,1,balanced,0.5594346523284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,1,balanced,0.6035626729329427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,1,balanced,0.6976213455200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,1,balanced,0.7815039952596029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,1,balanced,1.089296023050944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,1,balanced,1.2669973373413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,1,balanced,1.7718027432759602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,1,balanced,2.340336004892985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,1,balanced,3.202757199605306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,1,balanced,4.283077239990234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,power_law_1.2,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,power_law_1.2,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,power_law_1.2,0.030534398555755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,power_law_1.2,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,power_law_1.2,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,power_law_1.2,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,power_law_1.2,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,power_law_1.2,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,power_law_1.2,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,power_law_1.2,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,power_law_1.2,0.12154239416122437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,power_law_1.2,0.15365760326385497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.17359999418258668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.220633602142334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.2778239965438843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.33216640949249265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.3597568035125732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.359443211555481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.34716799259185793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.39231998920440675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.3887232065200806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.39496960639953616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.40302720069885256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.4278207778930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.443558406829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.45355520248413084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.48942079544067385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.5222335815429687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.5970943927764892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.6335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,power_law_1.01,0.7439167976379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,power_law_1.01,0.8433983802795411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,power_law_1.01,1.073408031463623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,power_law_1.01,1.2779199600219726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,power_law_1.01,1.775948715209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,power_law_1.01,2.2161855697631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,power_law_1.2,0.22422399520874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,power_law_1.2,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,power_law_1.2,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,power_law_1.2,0.07229440212249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,power_law_1.2,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,power_law_1.2,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,power_law_1.2,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,power_law_1.2,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,power_law_1.2,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,power_law_1.2,0.14001920223236083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,power_law_1.2,0.14759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,power_law_1.2,0.14986879825592042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,power_law_1.2,0.14298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,power_law_1.2,0.1561087965965271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,power_law_1.2,0.16563839912414552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,power_law_1.2,0.17200000286102296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,power_law_1.2,0.17980799674987794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,power_law_1.2,0.18730239868164061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,power_law_1.2,0.2127295970916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,power_law_1.2,0.2267712116241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,power_law_1.2,0.27822079658508303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,power_law_1.2,0.3209856033325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,power_law_1.2,0.4306623935699463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,power_law_1.2,0.49790081977844236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,power_law_1.2,0.695251178741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,power_law_1.2,0.81976318359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,power_law_1.2,1.148742389678955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,power_law_1.2,1.7001024246215821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,power_law_1.2,0.29085440635681153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,power_law_1.2,0.028064000606536865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,power_law_1.2,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,power_law_1.2,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,power_law_1.2,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,power_law_1.2,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,power_law_1.2,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,power_law_1.2,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,power_law_1.2,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,power_law_1.2,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,power_law_1.2,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,power_law_1.2,0.13640320301055908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,power_law_1.2,0.17052799463272095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,power_law_1.2,0.2239232063293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,power_law_1.2,0.28393599987030027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,balanced,0.08052800099054973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,balanced,0.08497066299120586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,balanced,0.08378666639328003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,balanced,0.08517332871754964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,balanced,0.08563199639320374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,balanced,0.08770133058230083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,balanced,0.09002133210500081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,balanced,0.08971200386683147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,balanced,0.10095466176668803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,balanced,0.10186133782068889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,balanced,0.10569066802660625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,balanced,0.11760000387827556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,balanced,0.1304266651471456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,balanced,0.15122133493423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,balanced,0.17094933986663818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,balanced,0.22714134057362875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,balanced,0.26502933104832965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,balanced,0.37884799639383954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,balanced,0.47838401794433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,balanced,0.6855733394622803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,balanced,0.9012320041656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,power_law_1.2,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,power_law_1.2,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,power_law_1.2,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,power_law_1.2,0.06218879818916321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,power_law_1.2,0.07703040242195129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,power_law_1.2,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,power_law_1.2,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,power_law_1.2,0.1268224000930786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,power_law_1.2,0.14117759466171265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,power_law_1.2,0.18769279718399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,power_law_1.2,0.2339711904525757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,power_law_1.2,0.3158911943435669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,power_law_1.2,0.4534592151641846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.11329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.13493759632110597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.18221440315246581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.19422719478607178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.1921728014945984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.19842560291290284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,power_law_1.2,0.5348991870880127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.2065664052963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.21755518913269042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.21906559467315673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.2267008066177368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.2453696012496948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.25806078910827634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.28592638969421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.299180793762207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.360531210899353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.4210368156433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.5165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.6121151924133301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,power_law_1.01,0.8023360252380372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.0506239891052247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,power_law_1.01,1.398323154449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,power_law_1.01,1.864352035522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,power_law_1.2,0.6902719974517822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,power_law_1.2,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,power_law_1.2,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,power_law_1.2,0.09630720019340515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,power_law_1.2,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,power_law_1.2,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,power_law_1.2,0.09339519739151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,power_law_1.2,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,power_law_1.2,0.0937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,power_law_1.2,0.09686400294303894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,power_law_1.2,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,power_law_1.2,0.09875199794769288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,power_law_1.2,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,power_law_1.2,0.10732799768447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,power_law_1.2,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,power_law_1.2,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,power_law_1.2,0.12310400009155273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,power_law_1.2,0.14254720211029054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,power_law_1.2,0.15573760271072387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,power_law_1.2,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,power_law_1.2,0.21772799491882325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,power_law_1.2,0.2646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,power_law_1.2,0.32663679122924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,power_law_1.2,0.4430079936981201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,power_law_1.2,0.6101503849029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,power_law_1.2,0.8002752304077149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,power_law_1.2,1.0637951850891114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.08959360122680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.09751039743423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.10242559909820556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.10896639823913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.12506879568099977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.1623103976249695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.18789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.22283520698547363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.2927232027053833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.37756800651550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.43617920875549315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.5858240127563477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,power_law_1.01,0.7906047821044921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.0977855682373048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,power_law_1.01,1.3742655754089355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,power_law_1.01,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,power_law_1.01,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,power_law_1.01,0.1204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,power_law_1.01,0.14713599681854247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,power_law_1.01,0.17163519859313964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,power_law_1.01,0.21937921047210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,power_law_1.01,0.2241152048110962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,power_law_1.01,0.2302720069885254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,power_law_1.01,0.23683838844299315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,power_law_1.01,0.24868481159210204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,power_law_1.01,0.25162880420684813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,power_law_1.01,0.2568831920623779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,power_law_1.01,0.26549758911132815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,power_law_1.01,0.2752511978149414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,power_law_1.01,0.28200318813323977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,power_law_1.01,0.2939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,power_law_1.01,0.32447359561920164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.01,0.3571968078613281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.01,0.41800317764282224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.01,0.4874559879302979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.01,0.620908784866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.01,0.7466368198394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.01,1.01014404296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.01,1.265881633758545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.01,1.7808639526367187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.01,2.2927743911743166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,power_law_1.01,0.0693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,power_law_1.01,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,power_law_1.01,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,power_law_1.01,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,power_law_1.01,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,power_law_1.01,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,power_law_1.01,0.09211519956588746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,power_law_1.01,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,power_law_1.01,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,power_law_1.01,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,power_law_1.01,0.09445120096206665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,power_law_1.01,0.09648000001907349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,power_law_1.01,0.10196479558944702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,power_law_1.01,0.10241279602050782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,power_law_1.01,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,power_law_1.01,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.01,0.11871360540390015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.01,0.13604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.01,0.16230399608612062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.01,0.19400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.01,0.23853440284729005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.01,0.31524479389190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.01,0.3900543928146362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.01,0.5795392036437989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.01,0.7519360065460206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,balanced,0.05774400134881338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,balanced,0.05952000121275584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,balanced,0.060378665725390114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,balanced,0.08700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,balanced,0.11753599842389424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,balanced,0.11935466527938843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,balanced,0.11963199575742085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,balanced,0.11984533071517944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,balanced,0.12062933047612508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,balanced,0.12294933199882507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,balanced,0.12543466687202454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,balanced,0.12370133399963379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,balanced,0.1281013290087382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,balanced,0.1329813301563263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,balanced,0.13555733362833658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,balanced,0.14414933323860168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,balanced,0.16101866960525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,balanced,0.1711946725845337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,balanced,0.20153067509333292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,balanced,0.2295893430709839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,balanced,0.2897813320159912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,balanced,0.3495519955952962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,balanced,0.5137226581573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,balanced,0.6398559808731079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,balanced,0.9266186555226644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,balanced,1.2070186932881672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,power_law_1.01,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,power_law_1.01,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,power_law_1.01,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,power_law_1.01,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,power_law_1.01,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,power_law_1.01,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,power_law_1.01,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,power_law_1.01,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,power_law_1.01,0.10506240129470826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,power_law_1.01,0.15073920488357545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,power_law_1.01,0.19528959989547728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,power_law_1.01,0.237062406539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,power_law_1.01,0.32641921043395994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,power_law_1.01,0.44092798233032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,power_law_1.2,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,power_law_1.2,0.06908800005912781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,power_law_1.2,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,power_law_1.2,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,power_law_1.2,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,power_law_1.2,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,power_law_1.2,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,power_law_1.2,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,power_law_1.2,0.09526399970054626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,power_law_1.2,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,power_law_1.2,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,power_law_1.2,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,power_law_1.2,0.15645439624786378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,power_law_1.2,0.1835968017578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,power_law_1.2,0.23595519065856935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,power_law_1.2,0.30052480697631834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,power_law_1.2,0.40687999725341795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,power_law_1.2,0.5047872066497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,power_law_1.2,0.7012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,power_law_1.2,0.937446403503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,power_law_1.01,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,power_law_1.01,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,power_law_1.01,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,power_law_1.01,0.07996799945831298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,power_law_1.01,0.10759040117263793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,power_law_1.01,0.10892159938812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,power_law_1.01,0.10942720174789429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,power_law_1.01,0.10554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,power_law_1.01,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,power_law_1.01,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,power_law_1.01,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,power_law_1.01,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,power_law_1.01,0.10721280574798583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,power_law_1.01,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,power_law_1.01,0.1144063949584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,power_law_1.01,0.11996159553527833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,power_law_1.01,0.11966079473495483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,power_law_1.01,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.01,0.18262399435043336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.01,0.1943295955657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.01,0.2545664072036743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.01,0.31719040870666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.01,0.4280128002166748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.01,0.5318784236907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.01,0.7619264125823975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.01,1.0287103652954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.14039679765701293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.24410240650177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.3371135950088501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.33517439365386964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.3377984046936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.31573760509490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.3001471996307373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.2804543972015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.3220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.32644479274749755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.2978559970855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.28915839195251464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.3251967906951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.2974720001220703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.32880001068115233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.33043200969696046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.33971199989318845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.34771199226379396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.3659136056900024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.4550271987915039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.44809598922729493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,power_law_1.01,0.5687168121337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,power_law_1.01,0.6383488178253174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,power_law_1.01,0.7855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,power_law_1.01,0.8164799690246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,power_law_1.01,1.0944704055786132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,power_law_1.01,1.3469504356384276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,power_law_1.01,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,power_law_1.01,0.04371199905872345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,power_law_1.01,0.07487360239028931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,power_law_1.01,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,power_law_1.01,0.10878080129623413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,power_law_1.01,0.10887680053710938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,power_law_1.01,0.1118016004562378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,power_law_1.01,0.1164031982421875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,power_law_1.01,0.11648639440536498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,power_law_1.01,0.13279999494552613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,power_law_1.01,0.13293440341949464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,power_law_1.01,0.1418944001197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,power_law_1.01,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,power_law_1.01,0.1655743956565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,power_law_1.01,0.19690879583358764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,power_law_1.01,0.22636160850524903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,power_law_1.01,0.29240319728851316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,power_law_1.01,0.33919999599456785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,power_law_1.01,0.45031042098999025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,power_law_1.01,0.5623936176300048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,power_law_1.01,0.7838719844818115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,power_law_1.01,1.0281536102294921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.2,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.2,0.053478401899337766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.2,0.05847679972648621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.2,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.2,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.2,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.2,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.2,0.08022400140762329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.2,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.2,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.2,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.2,0.16193920373916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.2,0.20280959606170654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.2,0.25192320346832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.2,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.2,0.38159999847412107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.2,0.47775998115539553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.2,0.6520383834838868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,power_law_1.01,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,power_law_1.01,0.05694079995155334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,power_law_1.01,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.01,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.01,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.01,0.10285439491271972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.01,0.1372480034828186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.01,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.01,0.23674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.01,0.3055552005767822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.01,0.4322048187255859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.01,0.5678592205047608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,balanced,0.12223466237386067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,balanced,0.12961600224177042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,balanced,0.12581866979599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,balanced,0.12552533547083536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,balanced,0.12429866194725037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,balanced,0.1264639993508657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,balanced,0.1317759950955709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,balanced,0.13193600376447043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,balanced,0.1359946628411611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,balanced,0.13541866342226663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,balanced,0.1343786617120107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,balanced,0.14061866203943887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,balanced,0.15225600202878317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,balanced,0.1527786652247111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,balanced,0.15500266353289285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,balanced,0.17495999733606973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,balanced,0.1807466745376587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,balanced,0.20616000890731812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,balanced,0.25498666365941364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,balanced,0.3171839912732442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,balanced,0.3866080045700073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,balanced,0.5425066550572714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,balanced,0.6723893483479818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,balanced,0.9531040191650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,balanced,1.2875786622365315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.11802239418029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.12046079635620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.1262336015701294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.13067519664764404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.13623679876327516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.14496639966964722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.15018240213394166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.16421760320663453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.19510400295257568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.2246335983276367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.2867136001586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.3483776092529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.47530879974365237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.5989632129669189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8508288383483886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.0917247772216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.5760831832885742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.0635135650634764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,power_law_1.2,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,power_law_1.2,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,power_law_1.2,0.07000960111618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,power_law_1.2,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,power_law_1.2,0.09611520171165466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.14750720262527467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.24026880264282227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.4339968204498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.545900821685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,power_law_1.2,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.878502368927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,power_law_1.2,0.09283840060234069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.2386879920959473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.3509568214416503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,power_law_1.2,0.09589120149612426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.6678207397460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,power_law_1.2,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,power_law_1.01,1.690412712097168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,power_law_1.01,1.7823360443115235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,power_law_1.2,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,power_law_1.2,0.11772160530090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,power_law_1.01,1.9535615921020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,power_law_1.2,0.1257599949836731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,power_law_1.2,0.14183679819107056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,power_law_1.01,1.8325439453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,power_law_1.2,0.1622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,power_law_1.2,0.18906240463256835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,power_law_1.01,1.8260736465454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,power_law_1.2,0.22062079906463622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,power_law_1.2,0.28881919384002686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,power_law_1.01,1.9751552581787108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,power_law_1.2,0.3608831882476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,power_law_1.01,2.046156883239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,power_law_1.2,0.469868803024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,power_law_1.2,0.6350336074829102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,power_law_1.01,2.007308769226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,power_law_1.2,0.9019136428833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.1694271087646486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,power_law_1.2,1.1859071731567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.1337087631225584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,power_law_1.01,2.3592384338378904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,power_law_1.01,2.741644859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,power_law_1.01,2.651206398010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,power_law_1.01,3.115782356262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,power_law_1.01,3.035148811340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,power_law_1.01,3.5198337554931642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,power_law_1.01,4.108179092407227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,power_law_1.01,5.031788635253906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,power_law_1.01,5.94458236694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.2,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.2,0.06903679966926575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.2,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.2,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.2,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.2,0.0749888002872467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.2,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.2,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.2,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.2,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.2,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.2,0.12189439535140992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.2,0.1402624011039734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.2,0.14586880207061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.2,0.18373119831085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.2,0.20602879524230958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.2,0.2746880054473877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.2,0.32016639709472655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.2,0.4635903835296631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.2,0.5954559803009033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.042534399032592776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,balanced,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,balanced,0.051914667089780174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,balanced,0.05372266471385956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,balanced,0.06241600215435028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,balanced,0.06251733501752217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,balanced,0.06450133522351582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,balanced,0.0784693310658137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,balanced,0.088837335507075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,balanced,0.10636267066001892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,balanced,0.13341333468755087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,balanced,0.15340800086657205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,balanced,0.2039146622021993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,balanced,0.2625173330307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,balanced,0.37218133608500165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,balanced,0.47348801294962567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.20999679565429688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.28380160331726073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.39823360443115235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.125164794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.1656448006629944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.23185279369354247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.2855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.47177600860595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.2,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.2,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.2,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.2,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.2,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.2,0.10481280088424683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.2,0.13703680038452148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.2,0.16316800117492675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.2,0.26131839752197267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.2,0.3325376033782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,power_law_1.01,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,power_law_1.01,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,power_law_1.01,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,power_law_1.01,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,power_law_1.01,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,power_law_1.01,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,power_law_1.01,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,power_law_1.01,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,power_law_1.01,0.0794431984424591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,power_law_1.01,0.0902463972568512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,power_law_1.01,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,power_law_1.01,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,power_law_1.01,0.11219199895858764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,power_law_1.01,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,power_law_1.01,0.14684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,power_law_1.01,0.17786879539489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,power_law_1.01,0.22093439102172852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,power_law_1.01,0.26905601024627684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,power_law_1.01,0.3736448049545288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,power_law_1.01,0.46136322021484377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,power_law_1.01,0.6874239921569825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,power_law_1.01,0.8627391815185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.1530751943588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.24906880855560304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.44425601959228517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.09244160056114196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.8211263656616211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.4621631622314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.1250240325927736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.093094402551651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.4596927642822264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.10280319452285766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.10576640367507935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.11813759803771973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.13597439527511596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.126688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.15920640230178834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.1853376030921936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.2538753509521485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.2336575984954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.263481593132019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.37047040462493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,power_law_1.2,0.4523200035095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.4213695526123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,power_law_1.2,0.6907904148101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,power_law_1.2,0.8742848396301269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.439411163330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.6763774871826174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,power_law_1.01,3.6595073699951173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,power_law_1.01,3.8283519744873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,power_law_1.01,3.9099456787109377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,power_law_1.01,4.14134407043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.075174331665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,power_law_1.01,4.301516723632813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,power_law_1.01,4.6190143585205075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,power_law_1.01,5.125094223022461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,power_law_1.01,5.1769664764404295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,power_law_1.01,6.064825439453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,power_law_1.01,5.800428771972657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,power_law_1.01,6.629017639160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,power_law_1.01,7.817798614501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,power_law_1.01,9.321459197998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,power_law_1.01,11.490726470947266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,power_law_1.2,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,power_law_1.2,0.11317119598388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,power_law_1.2,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,power_law_1.2,0.2659008026123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,power_law_1.2,0.3868160009384155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,power_law_1.2,0.5106175899505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,power_law_1.2,0.7497920036315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,power_law_1.2,0.7844160079956055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,power_law_1.2,0.8198783874511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,power_law_1.2,0.85283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,power_law_1.2,0.8880960464477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,power_law_1.2,0.9097855567932129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,power_law_1.2,0.9357824325561523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,power_law_1.2,0.9394816398620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,power_law_1.2,0.9563455581665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,power_law_1.2,1.0209152221679687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,power_law_1.2,1.0826111793518067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,power_law_1.2,1.1652671813964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,power_law_1.2,1.157267189025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,power_law_1.2,1.316652774810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,power_law_1.2,1.3208767890930175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,power_law_1.2,1.4911359786987304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,power_law_1.2,1.6294591903686524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,power_law_1.2,2.0065088272094727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,power_law_1.2,2.272159957885742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,power_law_1.2,2.8800447463989256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,power_law_1.2,3.4904510498046877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.2,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.2,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.2,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.2,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.2,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.2,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.2,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.2,0.1392639994621277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.2,0.1599679946899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.2,0.21848320960998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.2,0.23274879455566405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.2,0.3053247928619385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.2,0.37401599884033204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.2,0.5620863914489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.2,0.7120512008666993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,balanced,0.124208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,balanced,0.2121653358141581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,balanced,0.21287999550501505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,balanced,0.2169546683629354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,balanced,0.21407467126846313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,balanced,0.2187946637471517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,balanced,0.22582934300104776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,balanced,0.22489599386850992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,balanced,0.23086400826772055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,balanced,0.2319200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,balanced,0.2372693419456482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,balanced,0.24707732597986856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,balanced,0.25413866837819415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,balanced,0.26609599590301514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,balanced,0.2908639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,balanced,0.3209386666615804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,balanced,0.3898986577987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,balanced,0.4412320057551066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,balanced,0.5803573528925577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,balanced,0.6846559842427572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,balanced,0.9741546312967936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,balanced,1.223754644393921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,balanced,1.7647412618001301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,balanced,2.242304007212321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,power_law_1.2,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,power_law_1.2,0.04963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,power_law_1.2,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,power_law_1.2,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,power_law_1.2,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,power_law_1.2,0.05575039982795715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,power_law_1.2,0.05982720255851746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,power_law_1.2,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,power_law_1.2,0.08863360285758973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,power_law_1.2,0.09528319835662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,power_law_1.2,0.11726080179214478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,power_law_1.2,0.12721279859542847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,power_law_1.2,0.1702847957611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,power_law_1.2,0.20366721153259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,power_law_1.2,0.27966721057891847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,power_law_1.2,0.37110400199890137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,power_law_1.2,0.5546751976013183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,power_law_1.2,0.6815360069274903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,balanced,0.07137600084145863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,balanced,0.11121066411336263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,balanced,0.17449599504470825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,balanced,0.2981119950612386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,balanced,0.2988479932149251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,balanced,0.30109866460164386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,balanced,0.3027626673380534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,balanced,0.3016800085703532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,balanced,0.3023786743481954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,balanced,0.3059840003649394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,balanced,0.30667734146118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,balanced,0.3079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,balanced,0.31570667028427124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,balanced,0.31540799140930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,balanced,0.31945600112279254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,balanced,0.3288533290227254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,balanced,0.3389013210932414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,balanced,0.3551679849624634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,balanced,0.3837546507517497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,balanced,0.4333386818567912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,balanced,0.449402650197347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,balanced,0.5538080135981241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,balanced,0.6073973178863525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,balanced,0.8185226917266846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,balanced,0.9281760056813558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,balanced,0.07186133166154225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,balanced,0.10199999809265137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,balanced,0.14854400356610617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,balanced,0.20014933745066324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,balanced,0.20210667451222739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,balanced,0.2032960057258606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,balanced,0.20251200596491495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,balanced,0.203439990679423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,balanced,0.20610666275024414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,balanced,0.20879467328389487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,balanced,0.21296000480651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,balanced,0.21735467513402304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,balanced,0.2201546629269918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,balanced,0.22923199335734049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,balanced,0.24199465910593668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,balanced,0.2537440061569214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,balanced,0.28170667092005414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,balanced,0.30515732367833454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,balanced,0.35820265611012775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,balanced,0.42260801792144775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,balanced,0.5247999827067057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,balanced,0.6429920196533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,balanced,0.8579040368398031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,balanced,1.0889013608296711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.2,0.02908160090446472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.2,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.2,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.2,0.03889279961585999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.2,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.2,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.2,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.2,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.2,0.08657919764518737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.2,0.11019519567489625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.2,0.13660800457000732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.2,0.1846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.2,0.2421567916870117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.2,0.26874239444732667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,power_law_1.01,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,power_law_1.01,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,power_law_1.01,0.12280960083007812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,power_law_1.01,0.1606335997581482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,power_law_1.01,0.15824639797210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,power_law_1.01,0.16546560525894166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,power_law_1.01,0.17802239656448365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,power_law_1.01,0.16822400093078613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,power_law_1.01,0.17323520183563232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,power_law_1.01,0.18083200454711915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,power_law_1.01,0.17802239656448365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,power_law_1.01,0.1867967963218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,power_law_1.01,0.19649920463562012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,power_law_1.01,0.20293760299682617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,power_law_1.01,0.21890559196472167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,power_law_1.01,0.23621759414672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,power_law_1.01,0.25962879657745364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,power_law_1.01,0.31687679290771487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,power_law_1.01,0.3617791891098022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,power_law_1.01,0.45350399017333987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,power_law_1.01,0.543667221069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,power_law_1.01,0.6833983898162842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,power_law_1.01,0.861177635192871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,power_law_1.01,1.2073920249938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,power_law_1.01,1.531436824798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,power_law_1.01,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,power_law_1.01,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,power_law_1.01,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,power_law_1.01,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,power_law_1.01,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,power_law_1.01,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.01,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.01,0.14544639587402344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.01,0.1620352029800415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.01,0.22946560382843018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,balanced,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,balanced,0.03793066740036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,balanced,0.041290665666262306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,balanced,0.050698667764663696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,balanced,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,balanced,0.08921066919962566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,balanced,0.10421866178512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,balanced,0.1365653375784556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,balanced,0.1648373305797577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,balanced,0.23568532864252725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,balanced,0.30905065933863324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.01,0.32403199672698973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.01,0.4594751834869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.01,0.6171264171600341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,balanced,0.0823359986146291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,balanced,0.13218133648236594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,balanced,0.23828266064325967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,balanced,0.2350239952405294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,balanced,0.23700799544652304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,balanced,0.2372373342514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,balanced,0.23822933435440063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,balanced,0.23995200792948404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,balanced,0.2494879961013794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,balanced,0.25066133340199787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,balanced,0.24525866905848184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,balanced,0.24538666009902954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,balanced,0.24699733654658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,balanced,0.25569599866867065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,balanced,0.2717386682828267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,balanced,0.274725337823232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,balanced,0.27930132548014325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,balanced,0.32811200618743896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,balanced,0.33606934547424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,balanced,0.41570134957631427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,balanced,0.47865601380666095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,balanced,0.6526986757914225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,balanced,0.799232006072998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,balanced,1.1586720148722331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,balanced,1.4649386405944824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,balanced,2.184549331665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,balanced,2.866954803466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.01,0.9727999687194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.01,1.3359423637390138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,power_law_1.01,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,power_law_1.01,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,power_law_1.01,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,power_law_1.01,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,power_law_1.01,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,power_law_1.01,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,power_law_1.01,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,power_law_1.01,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,power_law_1.01,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,power_law_1.01,0.0951807975769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.01,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.01,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.01,0.2074496030807495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.01,0.2315135955810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.01,0.3357952117919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.01,0.4704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.01,0.6515071868896485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.01,1.0062335968017577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,balanced,0.06804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,balanced,0.09086400270462036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,balanced,0.24127999941507974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,balanced,0.43782933553059894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,balanced,0.8247733116149902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,balanced,0.8289706707000732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,balanced,0.8290399710337321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,balanced,0.8305173714955648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,balanced,0.8304693698883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,balanced,0.8347413539886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,balanced,0.8385759989420573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,balanced,0.8425599733988444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,balanced,0.8538186550140381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,balanced,0.8597119649251302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,balanced,0.8637706438700358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,balanced,0.8741493225097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,balanced,0.9006133079528809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,balanced,0.9253546396891276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,balanced,0.972879966100057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,balanced,1.0216960112253826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,balanced,1.1082613468170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,balanced,1.1652426719665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,balanced,1.4507999420166016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,balanced,1.5145972569783528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,balanced,2.233919938405355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,balanced,2.388879934946696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.09204480051994324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.10129280090332031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.11107840538024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.11132160425186158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.11840640306472779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.1213312029838562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.13063679933547973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.13249919414520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.15818239450454713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.16613759994506835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.2200063943862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.22761600017547606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.28216960430145266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.349727988243103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,power_law_1.2,0.5051136016845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,power_law_1.2,0.49762558937072754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.01,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.01,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.01,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.01,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.01,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.01,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.01,0.10110080242156982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.01,0.12753920555114745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.01,0.13941119909286498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.01,0.16687359809875488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.01,0.21441919803619386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.01,0.28149120807647704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.01,0.3326848030090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.01,0.5318592071533204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.01,0.5981632232666015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,power_law_1.2,0.7625728130340577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,power_law_1.2,0.921183967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,power_law_1.01,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,power_law_1.01,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,power_law_1.01,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,power_law_1.01,0.08435199856758117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,power_law_1.01,0.11480319499969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.01,0.13259520530700683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.01,0.1580415964126587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.01,0.2071295976638794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.01,0.31781759262084963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.01,0.39761919975280763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.01,0.6025472164154053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.01,0.7934783935546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.01,1.1867136001586913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.01,1.6076543807983399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,power_law_1.01,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,power_law_1.01,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,power_law_1.01,0.08856959939002991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,power_law_1.01,0.10407040119171143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,power_law_1.01,0.12246400117874146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,power_law_1.01,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,power_law_1.01,0.15096319913864137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,power_law_1.01,0.15806080102920533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,power_law_1.01,0.1578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,power_law_1.01,0.15992319583892822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,power_law_1.01,0.1690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,power_law_1.01,0.17149440050125123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,power_law_1.01,0.1831104040145874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,power_law_1.01,0.20052480697631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,power_law_1.01,0.21130878925323487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,power_law_1.01,0.21941759586334228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,power_law_1.01,0.2497472047805786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,power_law_1.01,0.2731584072113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,power_law_1.01,0.3261503934860229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,power_law_1.01,0.39003520011901854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,power_law_1.01,0.5226816177368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,power_law_1.01,0.6261312007904053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,power_law_1.01,0.8331392288208008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,power_law_1.01,1.0168255805969237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,power_law_1.01,1.4321536064147948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,power_law_1.01,1.8413888931274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.11012480258941651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.10392960309982299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.10927360057830811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.09998720288276672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.11307519674301147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.11018240451812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.12660479545593262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.12920960187911987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.13997440338134765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.15359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.15886720418930053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.19121919870376586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.21979520320892335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.2854144096374512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.34426240921020507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.459168004989624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.5849728107452392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.7943679809570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.1670080184936524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.049830400943756105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.07703040242195129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.01,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.01,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.01,0.029183998703956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.01,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.01,0.029491201043128967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.01,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.01,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.01,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.01,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.01,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.01,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.01,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.01,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.01,0.05808640122413635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.01,0.1264896035194397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.01,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.01,0.19514240026474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.01,0.22267520427703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.22200319766998292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.1472767949104309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.17840640544891356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.2509887933731079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.2682368040084839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.27319679260253904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.28828160762786864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.2873087882995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.29487359523773193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.31288321018218995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.3115648031234741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.3251967906951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.33281919956207273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.3397183895111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.35676159858703616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.3835839986801147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.4156223773956299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.5085951805114746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.5558976173400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,power_law_1.2,0.6701695919036865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,power_law_1.2,0.7721024036407471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,power_law_1.2,0.9912704467773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,power_law_1.2,1.1753855705261231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.2948992013931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,power_law_1.2,1.7914751052856446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,power_law_1.2,2.1658496856689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,balanced,0.02959999938805898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,balanced,0.05629333357016245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,balanced,0.05969599882761637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,balanced,0.05861866474151611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,balanced,0.0590826670328776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,balanced,0.06192000210285187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,balanced,0.06165333092212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,balanced,0.065461332599322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,balanced,0.0718506673971812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,balanced,0.0701386680205663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,balanced,0.07342400153477986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,balanced,0.08016000191370647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,balanced,0.09461866815884908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,balanced,0.105103999376297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,balanced,0.11162133018175761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,balanced,0.1418560047944387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,balanced,0.1681706706682841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,balanced,0.2319200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,balanced,0.27769599358240765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,balanced,0.40037333965301514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,balanced,0.5168746709823608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,balanced,0.03384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,balanced,0.03793066740036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,balanced,0.03711466739575068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,balanced,0.04271466533342997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,balanced,0.04192000130812327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,balanced,0.05843733251094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,balanced,0.06943466762701671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,balanced,0.11251733700434367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,balanced,0.1323466698328654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,balanced,0.1874879995981852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,balanced,0.22595200935999551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.07291520237922669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,balanced,0.08012799918651581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,balanced,0.08110933502515157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,balanced,0.07963733375072479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.09431679844856262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,balanced,0.08434133728345235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,balanced,0.08574933807055156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,balanced,0.0846453309059143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,balanced,0.08450133601824443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,balanced,0.08824533224105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,balanced,0.09009066224098206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,balanced,0.08885332942008972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,balanced,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,balanced,0.1019040048122406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,balanced,0.10162666440010071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,balanced,0.10739733775456746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,balanced,0.13005333145459494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,balanced,0.13366400202115378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,balanced,0.16335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,balanced,0.19205333789189658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,balanced,0.23813333113988241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,balanced,0.28541332483291626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,balanced,0.3986186583836873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,balanced,0.5050133466720581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,balanced,0.7148373126983643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.1165887951850891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,balanced,0.952245314915975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.1282240033149719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.16850559711456298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.19809919595718384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.24600958824157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.325164794921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.450867223739624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,power_law_1.01,0.5923200130462647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,power_law_1.01,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,power_law_1.01,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,power_law_1.01,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,power_law_1.01,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,power_law_1.01,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,power_law_1.01,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,power_law_1.01,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,power_law_1.01,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,power_law_1.01,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,power_law_1.01,0.1503424048423767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,power_law_1.01,0.1968384027481079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,power_law_1.01,0.24910080432891846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,power_law_1.01,0.36246399879455565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,power_law_1.01,0.44330878257751466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,power_law_1.01,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,power_law_1.01,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,power_law_1.01,0.06724479794502258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,power_law_1.01,0.08064000010490417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,power_law_1.01,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,power_law_1.01,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,power_law_1.01,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,power_law_1.01,0.12191359996795655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,power_law_1.01,0.12755199670791625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,power_law_1.01,0.1613055944442749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,power_law_1.01,0.19256319999694824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,power_law_1.01,0.24871039390563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,power_law_1.01,0.31427841186523436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,power_law_1.01,0.4435455799102783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,power_law_1.01,0.5383999824523926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,power_law_1.01,0.7833856105804443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,power_law_1.01,1.0118016242980956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.05934079885482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06146559715270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.1211583971977234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.1578112006187439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.18129279613494872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.28255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,power_law_1.01,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,power_law_1.01,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,power_law_1.01,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,power_law_1.01,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,power_law_1.01,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,power_law_1.01,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,power_law_1.01,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,power_law_1.01,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,power_law_1.01,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,power_law_1.01,0.11667200326919555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,power_law_1.01,0.1401792049407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.01,0.16165759563446044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.36012160778045654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.01,0.23264000415802003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.01,0.28494720458984374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.01,0.39416959285736086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.01,0.520307207107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.01,0.7579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.5040703773498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.01,1.0624128341674806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.6686975955963135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.01,1.6194368362426759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,power_law_1.2,0.025740799307823182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,power_law_1.2,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,power_law_1.2,0.027001601457595826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,power_law_1.2,0.028575998544692994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,power_law_1.2,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,power_law_1.2,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,power_law_1.2,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,power_law_1.2,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.06952959895133973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.0769216001033783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.11591039896011353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.15034879446029664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.01,1.9076223373413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,1.0023232460021974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.21571838855743408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,balanced,0.05650666852792104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,balanced,0.06132799883683523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,balanced,0.06462400158246358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,balanced,0.06716800232728322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,balanced,0.06938666601975758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,balanced,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,balanced,0.06646400193373363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,balanced,0.07300266623497009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,balanced,0.07668266693751018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,balanced,0.08331199983755748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.08694932858149211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10342933734258015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.11729600032170613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.1425226628780365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.16607466340065002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.21453332901000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.2568426728248596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.35808531443277997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.4473493496576945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.27761919498443605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,1.3259967803955077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,power_law_1.2,0.34012160301208494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,balanced,0.02740799884001414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,balanced,0.02737066646416982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,balanced,0.031023999055226643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,balanced,0.03928533444801966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,balanced,0.06814933319886525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,balanced,0.08482133348782857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,balanced,0.1030613382657369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,balanced,0.12523200114568075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,balanced,0.04062400013208389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,balanced,0.045066664616266884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,balanced,0.054570664962132774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,balanced,0.05667733152707418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,balanced,0.057018667459487915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,balanced,0.05962666869163513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,balanced,0.06275733311971028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,balanced,0.07839466631412506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,balanced,0.09345600008964539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,balanced,0.10568533341089885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,balanced,0.14711999893188477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,balanced,0.17778666814168295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,balanced,0.24395199616750082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,balanced,0.2930720051129659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,balanced,0.4106026490529378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,balanced,0.5272106726964315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.08175359964370728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.08344320058822632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.09051520228385926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.09244160056114196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.09875839948654175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.10650880336761474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.11244800090789794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.1255231976509094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.150271999835968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.17160320281982422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.21831679344177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.2627392053604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.3605247974395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.4631807804107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.6647488117218018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.862342357635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,1.270246410369873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.6549184799194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.07570559978485107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.11326719522476196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.14431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.17621760368347167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.24866559505462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.3317312002182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,balanced,0.056549335519472756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,balanced,0.08810133735338847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,balanced,0.08939733107884724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,balanced,0.0929813285668691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,balanced,0.09335999687512715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,balanced,0.09524800380071004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,balanced,0.09737599889437358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,balanced,0.10591466228167216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,balanced,0.10459199547767639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,balanced,0.10838400324185689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,balanced,0.11952533324559529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,balanced,0.12643200159072876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,balanced,0.14710399508476257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,balanced,0.16665066281954447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,balanced,0.20342934131622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,balanced,0.23888534307479858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,balanced,0.33275200923283893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,balanced,0.4081813494364421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,balanced,0.5952159961064657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,balanced,0.7647466659545898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.061382400989532473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.0826367974281311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.12208640575408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.14357119798660278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.1865023970603943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.2,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.2,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.2,0.09167360067367554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.2,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.2,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.2,0.10452480316162109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.2,0.10309760570526123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.2,0.10536960363388062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.2,0.10967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.2,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.2,0.11820800304412842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.2,0.118886399269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.2,0.12940160036087037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.2,0.13199360370635987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.2,0.13852800130844117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.2,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.2,0.1773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.2,0.2248447895050049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.2,0.2728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.2,0.33736319541931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.2,0.368831992149353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.2,0.49334402084350587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.23291521072387694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.2,0.6063680171966552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.2,0.8806912422180175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.2,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.2,0.09388160109519958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.2,1.1678336143493653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.2,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.2,0.1038591980934143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.2,0.1047551989555359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.2,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.2,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.2,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.2,0.10138880014419556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.2,0.10467840433120727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.2,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.2,0.11008640527725219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.2,0.11471999883651733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.2,0.1251904010772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.2,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.2,0.12893439531326295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.2,0.14716800451278686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.2,0.167468798160553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.2,0.2041856050491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.2,0.2397183895111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.2,0.30662400722503663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.2,0.3276479959487915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.2,0.45650558471679686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.2,0.5499648094177246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.2,0.7912447929382325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.2,0.9947135925292969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.3554111957550049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.40147838592529295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,power_law_1.01,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,power_law_1.01,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,power_law_1.01,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,power_law_1.01,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,power_law_1.01,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,power_law_1.01,0.06861439943313599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,power_law_1.01,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,power_law_1.01,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,power_law_1.01,0.07925119996070862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,power_law_1.01,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,power_law_1.01,0.08787840008735656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,power_law_1.01,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,power_law_1.01,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,power_law_1.01,0.15189759731292723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,power_law_1.01,0.1748800039291382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,power_law_1.01,0.22446079254150392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,power_law_1.01,0.28264319896698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,power_law_1.01,0.3790208101272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,power_law_1.01,0.4992063999176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,power_law_1.01,0.6826816082000733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,power_law_1.01,0.8791168212890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,power_law_1.2,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,power_law_1.2,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,power_law_1.2,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,power_law_1.2,0.04531840085983276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,power_law_1.2,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,power_law_1.2,0.07132160067558288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.2,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.2,0.1102463960647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,balanced,0.058730666836102806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,balanced,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.2,0.1306496024131775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,balanced,0.0599839985370636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,balanced,0.0726453314224879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,balanced,0.08973333239555359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,balanced,0.09909866253534953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,balanced,0.10139733552932739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,balanced,0.09954133629798889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,balanced,0.10177066922187805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,balanced,0.10354666908582051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,balanced,0.10521066188812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,balanced,0.10677867134412129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,balanced,0.10864532987276714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,balanced,0.11852266391118367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,balanced,0.11949333548545837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,balanced,0.12506666779518127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,balanced,0.1301866670449575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,balanced,0.13823466499646506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,balanced,0.1564906636873881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,balanced,0.16806934277216592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,balanced,0.20893333355585733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,balanced,0.23643199602762857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,balanced,0.29392532507578534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,balanced,0.3709119955698649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,balanced,0.4858666658401489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,balanced,0.625653346379598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.2,0.17793279886245728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.2,0.2087104082107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.2,0.32152960300445554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.2,0.40723838806152346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.11580159664154052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.13495680093765258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.1708351969718933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.2135551929473877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.2539072036743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.3604032039642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.3740927934646606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.38473598957061766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.39864959716796877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.4059584140777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.4136832237243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.42995200157165525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.4318272113800049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.45372161865234373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.45326719284057615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.467852783203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.4908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.5485631942749023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.6146751880645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.7259136199951172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.8363200187683105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,1.0869183540344238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.3368063926696778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.8498495101928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,2.3595903396606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,3.384320068359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,4.409247970581054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,power_law_1.2,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,power_law_1.2,0.08154240250587463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,power_law_1.2,0.08932480216026306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,power_law_1.2,0.08976640105247498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,power_law_1.2,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,power_law_1.2,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,power_law_1.2,0.09036800265312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,power_law_1.2,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,power_law_1.2,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,power_law_1.2,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,power_law_1.2,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,power_law_1.2,0.10514559745788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,power_law_1.2,0.10476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,power_law_1.2,0.12046719789505005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.2,0.1359488010406494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.2,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.2,0.18536319732666015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.2,0.237939190864563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.2,0.3043584108352661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.2,0.4648831844329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.2,0.551475191116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.2,0.838259220123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.2,1.1395263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.2,0.7427519798278809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.2,0.9883647918701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.13752959966659545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.14863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.15915520191192628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.1575744032859802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.15994880199432374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.17420799732208253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.1739583969116211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.1745344042778015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.18445440530776977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.17823359966278077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.18513280153274536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.19497599601745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.197542405128479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.2106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.23880319595336913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.2392319917678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.2641535997390747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.3014784097671509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.36999680995941164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,power_law_1.01,0.4307392120361328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,power_law_1.01,0.551206398010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,power_law_1.01,0.7149184226989747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,power_law_1.01,0.8930432319641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,power_law_1.01,1.0961471557617188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,power_law_1.2,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,power_law_1.2,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,power_law_1.2,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,power_law_1.2,0.07266560196876526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,power_law_1.2,0.07196159958839417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,power_law_1.2,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,power_law_1.2,0.08038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,power_law_1.2,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,power_law_1.2,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,power_law_1.2,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,power_law_1.2,0.10599679946899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,power_law_1.2,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,power_law_1.2,0.12664320468902587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,power_law_1.2,0.13756799697875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,power_law_1.2,0.17439359426498413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,power_law_1.2,0.2034816026687622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,power_law_1.2,0.2589888095855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,power_law_1.2,0.3342144012451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,power_law_1.2,0.46604161262512206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,power_law_1.2,0.5502848148345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,balanced,0.0890880028406779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,balanced,0.09314666191736858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,balanced,0.10709866881370544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,balanced,0.13010666767756143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,balanced,0.16586132844289145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,balanced,0.1678719917933146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,balanced,0.17122133572896323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,balanced,0.17087467511494955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,balanced,0.17374932765960693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,2,balanced,0.17393600940704346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,2,balanced,0.17734400431315103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,2,balanced,0.17798932393391928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,2,balanced,0.18224000930786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,2,balanced,0.18121600151062012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,2,balanced,0.1844373345375061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,2,balanced,0.19183999300003052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,2,balanced,0.2064639925956726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,2,balanced,0.21627734104792276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,2,balanced,0.24952000379562378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,2,balanced,0.2772853374481201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,2,balanced,0.33777066071828205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,2,balanced,0.4002186854680379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,2,balanced,0.5658560196558634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,2,balanced,0.692512035369873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,2,balanced,0.9927573204040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,2,balanced,1.2951839764912922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.2,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.2,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.2,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.2,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.2,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.2,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.2,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.2,0.09726079702377319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.2,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.2,0.1148095965385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.2,0.13263360261917115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.2,0.14424320459365844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.2,0.18335360288619995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.2,0.1975551962852478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.2,0.2592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.2,0.3033407926559448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.2,0.4107647895812988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.2,0.53089280128479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.01,0.11713919639587403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.01,0.18139519691467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.01,0.18042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.01,0.24679040908813477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.01,0.3072256088256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.01,0.4078335762023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.01,0.38311679363250734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.01,0.39857280254364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.01,0.3930880069732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.01,0.40929279327392576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.01,0.4052159786224365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.01,0.4324160099029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.01,0.42865281105041503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.01,0.4302207946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.01,0.4537792205810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.01,0.4547584056854248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.01,0.49716482162475584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.01,0.543552017211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.01,0.5501567840576171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.01,0.6511680126190186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.01,0.7522751808166503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.01,0.929964828491211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.01,0.9603455543518067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.01,1.1643648147583008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.01,1.4138367652893067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.01,1.8891584396362304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.01,2.2544319152832033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,power_law_1.01,0.056908798217773435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,power_law_1.01,0.07330560088157653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,power_law_1.01,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,power_law_1.01,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,power_law_1.01,0.09900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,power_law_1.01,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,power_law_1.01,0.10395519733428955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,power_law_1.01,0.10282880067825317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,power_law_1.01,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,power_law_1.01,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,power_law_1.01,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,power_law_1.01,0.11148799657821655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,power_law_1.01,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,power_law_1.01,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,power_law_1.01,0.11873279809951783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,power_law_1.01,0.13109760284423827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,power_law_1.01,0.1368064045906067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,power_law_1.01,0.15932159423828124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,power_law_1.01,0.1757696032524109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,power_law_1.01,0.22469758987426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,power_law_1.01,0.22624640464782714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,power_law_1.01,0.30370559692382815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,power_law_1.01,0.36498560905456545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,power_law_1.01,0.4989312171936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,power_law_1.01,0.6249407768249512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,power_law_1.2,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,power_law_1.2,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,power_law_1.2,0.05851519703865051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,power_law_1.2,0.06261759996414185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,power_law_1.2,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,power_law_1.2,0.06264960169792175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,power_law_1.2,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,power_law_1.2,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,power_law_1.2,0.09989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,power_law_1.2,0.12377599477767945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,power_law_1.2,0.20054399967193604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,power_law_1.2,0.23704960346221923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,power_law_1.2,0.34203519821166994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,power_law_1.2,0.4325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,power_law_1.2,0.5970751762390136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,power_law_1.2,0.8307071685791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,balanced,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,balanced,0.08869333068529765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,balanced,0.12310399611790974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,balanced,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,balanced,0.18300267060597739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,balanced,0.18528532981872559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,balanced,0.18694400787353516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,balanced,0.1846346656481425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,balanced,0.18941332896550497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,balanced,0.19079466660817465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,balanced,0.19230933984120688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,balanced,0.19555199146270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,balanced,0.20288532972335815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,balanced,0.20595200856526694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,balanced,0.21179733673731485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,balanced,0.22929600874582926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.24265066782633463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.27457600831985474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.30703999598821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.3555786609649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.4041653474171956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.56004265944163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.6485919952392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,balanced,0.9233813285827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.1575733025868733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.2,0.10275839567184449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.2,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.2,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.2,0.13929599523544312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.2,0.15136640071868895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.2,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.2,0.14863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.2,0.1385856032371521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.2,0.14248960018157958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.2,0.14915839433670045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.2,0.15219839811325073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.2,0.15716480016708373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.2,0.1625216007232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.2,0.1739456057548523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.2,0.1744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.2,0.1859071969985962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.2,0.2075711965560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.2,0.23400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.2,0.2825920104980469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.2,0.30382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.2,0.401094388961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.2,0.41728639602661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.2,0.556544017791748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.2,0.7183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.2,1.0038528442382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.11176960468292237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.15301120281219482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.1614848017692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.1518272042274475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.15975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.13224320411682128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.15797120332717896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.172326397895813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.17326079607009887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.23372159004211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.23502719402313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,power_law_1.2,0.34438400268554686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,power_law_1.2,0.4211264133453369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.2,1.3009920120239258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,power_law_1.2,0.566048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,power_law_1.2,0.7730432033538819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,power_law_1.2,1.077184009552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,power_law_1.2,1.461740779876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,power_law_1.2,2.2180416107177736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.025753599405288697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.02776319980621338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.027833598852157592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.147980797290802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.1965824007987976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,power_law_1.2,2.858291244506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,power_law_1.01,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,power_law_1.01,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,power_law_1.01,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,power_law_1.01,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,power_law_1.01,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,power_law_1.01,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,power_law_1.01,0.08170880079269409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,power_law_1.01,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,power_law_1.01,0.10426880121231079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,power_law_1.01,0.1354688048362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.266048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,power_law_1.01,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,power_law_1.01,0.22576639652252198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,power_law_1.01,0.2918207883834839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,power_law_1.01,0.4039487838745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,power_law_1.01,0.512883186340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.29180800914764404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,power_law_1.01,0.7046144008636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,power_law_1.01,0.8900544166564941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,balanced,0.08822932839393616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,balanced,0.14618133505185446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,balanced,0.1384053329626719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,balanced,0.13884799679120383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,balanced,0.13823999961217245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,balanced,0.13928000132242838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,balanced,0.14160533746083578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,balanced,0.1476533313592275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,balanced,0.1488053301970164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,balanced,0.1441333293914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,balanced,0.14456533392270407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,balanced,0.14433599511782327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,balanced,0.15121600031852722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,balanced,0.16148266196250916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,balanced,0.1620266636212667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,balanced,0.16710400581359863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,balanced,0.1871946652730306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,balanced,0.20595733324686685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,balanced,0.25810132424036664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,balanced,0.2650453249613444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,balanced,0.3545440038045247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,balanced,0.4240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,balanced,0.6092480023701986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,balanced,0.7952853043874105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,balanced,1.1647520065307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,balanced,1.4693387349446614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.07242879867553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.09943680167198181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.12034560441970825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.17764480113983155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.18687360286712645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.19203200340270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.19785599708557128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.20112640857696534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.20584959983825685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.21326720714569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.21826560497283937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.21895039081573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.23064959049224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.24011518955230712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.25262720584869386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.2768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.3098047971725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.36951680183410646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.4191296100616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.5245247840881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.6271488189697265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,power_law_1.01,0.8457856178283691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.0303104400634766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,power_law_1.01,1.4438015937805175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,power_law_1.01,1.8390975952148438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,power_law_1.2,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,power_law_1.2,0.047686401009559634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,power_law_1.2,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,power_law_1.2,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,power_law_1.2,0.06113920211791992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,power_law_1.2,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,power_law_1.2,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,power_law_1.2,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.2,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.2,0.11912319660186768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.2,0.2151616096496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.2,0.31331839561462405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.2,0.39838080406188964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.2,0.7333055973052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.2,0.9620479583740235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.09925119876861573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.10307199954986572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.10833920240402221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.11584639549255371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.11802879571914673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.13784960508346558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.1481152057647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.16165759563446044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.17916799783706666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.20422399044036865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.27234559059143065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.30675199031829836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.39237759113311765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.5147583961486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,0.6442880153656005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.0256768226623536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.6663167953491211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,1.9980928421020507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,power_law_1.2,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,power_law_1.2,0.1045632004737854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,power_law_1.2,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,power_law_1.2,0.10750720500946045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,power_law_1.2,0.10664960145950317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,power_law_1.2,0.1205183982849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,power_law_1.2,0.1268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,power_law_1.2,0.1355455994606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,power_law_1.2,0.14534399509429932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,power_law_1.2,0.153164803981781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,power_law_1.2,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,power_law_1.2,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,power_law_1.2,0.2128063917160034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,power_law_1.2,0.2662528038024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,power_law_1.2,0.3285696029663086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,power_law_1.2,0.419155216217041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,power_law_1.2,0.5216832160949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,power_law_1.2,0.7256959915161133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,power_law_1.2,0.9745216369628906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,power_law_1.2,1.4009087562561036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,power_law_1.2,1.7077247619628906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.17047040462493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.17668479681015015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.2546880006790161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.40023040771484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.3904639959335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.3917184114456177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.397542405128479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.4120704174041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.41379199028015134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.42632322311401366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.44214401245117185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.46554880142211913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.48700799942016604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.4996352195739746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,power_law_1.2,0.5092031955718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,power_law_1.2,0.558022403717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,power_law_1.2,0.6121664047241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,power_law_1.2,0.6860608100891114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,power_law_1.2,0.78024320602417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,power_law_1.2,0.9174464225769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.062028801441192626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.06909440159797668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.1229375958442688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.275603199005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.35050239562988283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.5171008110046387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.7121984004974365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,power_law_1.2,1.18471040725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,power_law_1.2,1.4884223937988281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,power_law_1.2,1.9672767639160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,power_law_1.2,2.834476852416992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,2,balanced,0.11239467064539592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,2,balanced,0.11739733815193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,2,balanced,0.14838932951291403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,2,balanced,0.20888533194859824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,2,balanced,0.3372746706008911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,2,balanced,0.3931573232014974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,2,balanced,0.39610131581624347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,2,balanced,0.3992533286412557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,2,balanced,0.3985546827316284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,2,balanced,0.4027413527170817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,2,balanced,0.4047306776046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,2,balanced,0.4076799949010213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,2,balanced,0.4121919870376587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,2,balanced,0.41838932037353516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,2,balanced,0.4203999837239583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,2,balanced,0.42521599928538006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,2,balanced,0.4330506722132365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,2,balanced,0.4575680096944173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,2,balanced,0.4803146521250407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,2,balanced,0.5331626733144125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,2,balanced,0.5844106674194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,2,balanced,0.7888800303141276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,2,balanced,0.8906986713409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,2,balanced,1.230570634206136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,2,balanced,1.6151893933614094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,2,balanced,2.1656319300333657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,2,balanced,2.894789377848307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,power_law_1.2,3.5810688018798826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,power_law_1.01,0.042419201135635375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,power_law_1.01,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,power_law_1.01,0.037190398573875426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,power_law_1.01,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,power_law_1.01,0.14625920057296754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,power_law_1.01,0.15944960117340087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,power_law_1.01,0.22865281105041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,power_law_1.01,0.28295040130615234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,power_law_1.01,0.3842367887496948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,power_law_1.01,0.5357632160186767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,power_law_1.2,6.384492874145508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,balanced,0.039962666730086006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,balanced,0.05418133238951365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,balanced,0.07287466526031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,balanced,0.10937066872914632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,balanced,0.17132800817489624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,balanced,0.17222400506337485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,balanced,0.17380799849828085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,balanced,0.176362673441569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,balanced,0.17501866817474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,balanced,0.17543999354044595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,balanced,0.17706133921941122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,balanced,0.1776533325513204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,balanced,0.17770665884017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,balanced,0.18367467323939005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,balanced,0.18438933293024698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,balanced,0.18784000476201376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,balanced,0.1948853333791097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,balanced,0.20067733526229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,balanced,0.21381332476933798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,balanced,0.22641066710154215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,balanced,0.2609279950459798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,balanced,0.27552000681559247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,balanced,0.33814934889475506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,balanced,0.37853864828745526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,balanced,0.5082026720046997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,balanced,0.59497598807017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,power_law_1.2,7.849664306640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,balanced,0.02956799914439519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,balanced,0.04002666721741358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,balanced,0.05228800078233083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,balanced,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,balanced,0.11334400375684102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,balanced,0.15862933794657388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,balanced,0.22851733366648355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,balanced,0.26660799980163574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,balanced,0.4019999901453654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,balanced,0.4952586491902669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,power_law_1.2,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,power_law_1.2,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,power_law_1.2,0.0773248016834259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,power_law_1.2,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,power_law_1.2,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,power_law_1.2,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,power_law_1.2,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,power_law_1.2,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,power_law_1.2,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,power_law_1.2,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,power_law_1.2,0.10113279819488526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,power_law_1.2,0.10972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,power_law_1.2,0.13456640243530274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,power_law_1.2,0.16836479902267457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,power_law_1.2,0.17023359537124633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,power_law_1.2,0.21583359241485595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,power_law_1.2,0.2672640085220337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,power_law_1.2,0.3867136001586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,power_law_1.2,0.4844543933868408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,power_law_1.2,0.6902207851409912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,power_law_1.2,0.0942080020904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,power_law_1.2,0.12055679559707641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,power_law_1.2,0.1375488042831421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,power_law_1.2,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,power_law_1.2,0.18742400407791138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,power_law_1.2,0.18250880241394044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,power_law_1.2,0.19652479887008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,power_law_1.2,0.17832959890365602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,power_law_1.2,0.18179839849472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,power_law_1.2,0.19007359743118285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,power_law_1.2,0.17804800271987914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,power_law_1.2,0.16960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,power_law_1.2,0.19111039638519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,power_law_1.2,0.20678400993347168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,power_law_1.2,0.19965440034866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,power_law_1.2,0.2294912099838257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,power_law_1.2,0.2366015911102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,power_law_1.2,0.26831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,power_law_1.2,0.3216383934020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,power_law_1.2,0.3401983976364136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,power_law_1.2,0.4295231819152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,power_law_1.2,0.5070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,power_law_1.2,0.6554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,power_law_1.2,0.8253696441650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,power_law_1.2,1.128326416015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,power_law_1.2,1.4980928421020507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,power_law_1.01,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,power_law_1.01,0.05130239725112915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,power_law_1.01,0.0555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,power_law_1.01,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,power_law_1.01,0.06176000237464905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,power_law_1.01,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,power_law_1.01,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,power_law_1.01,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,power_law_1.01,0.08385279774665833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,power_law_1.01,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,power_law_1.01,0.1568127989768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,power_law_1.01,0.2026815891265869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,power_law_1.01,0.23914239406585694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,power_law_1.01,0.324947190284729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,power_law_1.01,0.4230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,power_law_1.01,0.6032192230224609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,power_law_1.01,0.8118399620056153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,power_law_1.2,0.8352959632873536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,power_law_1.01,0.02869119942188263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,power_law_1.01,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,power_law_1.01,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,power_law_1.01,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,power_law_1.01,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,power_law_1.01,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,power_law_1.01,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,power_law_1.01,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,power_law_1.01,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,power_law_1.01,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,power_law_1.01,0.08696960210800171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,power_law_1.01,0.09509119987487794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,power_law_1.01,0.11703679561614991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,power_law_1.01,0.13080960512161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,power_law_1.01,0.17103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,power_law_1.01,0.21303679943084716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,power_law_1.01,0.2957312107086182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,power_law_1.01,0.38188800811767576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,power_law_1.01,0.5469056129455566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,power_law_1.01,0.7083648204803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.01,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.01,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.01,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.01,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.01,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.01,0.05072000026702881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.01,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.01,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.01,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.01,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.01,0.13838080167770386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.01,0.18639999628067017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.01,0.2604351997375488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.10512640476226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.13229440450668334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.18673280477523804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.24159998893737794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.3108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.43603200912475587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.4491583824157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4721856117248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.47536001205444334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.5061247825622559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5096511840820312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.519814395904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5464255809783936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.5715904235839844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.6081791877746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6383872032165527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7096384048461915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.7669695854187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.9096960067749024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.9477631568908691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.1584063529968263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.3804800033569335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,1.7859840393066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,2.19564151763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,2.9882240295410156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,3.8086273193359377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,power_law_1.2,0.14911359548568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,power_law_1.2,0.2419071912765503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,power_law_1.2,0.3950655937194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,power_law_1.2,0.5793920040130616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,power_law_1.2,0.7659008026123046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,power_law_1.2,1.1819968223571777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,power_law_1.2,1.2104960441589356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,power_law_1.2,1.2190400123596192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,power_law_1.2,1.2865983963012695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,power_law_1.2,1.376095962524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,power_law_1.2,1.402899169921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,power_law_1.2,1.450937557220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,power_law_1.2,1.4532992362976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,power_law_1.2,1.5079232215881349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,power_law_1.2,1.548095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,power_law_1.2,1.6038080215454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,power_law_1.2,1.7246784210205077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,power_law_1.2,1.8337535858154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,power_law_1.2,2.042310333251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,power_law_1.2,2.0972736358642576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,power_law_1.2,2.4339967727661134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,power_law_1.2,2.649843215942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,power_law_1.2,3.3116031646728517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,power_law_1.2,3.7748798370361327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,power_law_1.2,4.856172943115235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,power_law_1.2,5.992748641967774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.08963199853897094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.09916160106658936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.12774399518966675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.14134399890899657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.16996480226516725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.1819264054298401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.2313983917236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.3042815923690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.4100224018096924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.5060736179351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.7163392066955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,0.895628833770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,power_law_1.2,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,power_law_1.2,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,power_law_1.2,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,power_law_1.2,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,power_law_1.2,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,power_law_1.2,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,power_law_1.2,0.06588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,power_law_1.2,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,power_law_1.2,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,power_law_1.2,0.15620479583740235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,power_law_1.2,0.1985535979270935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,power_law_1.2,0.28776960372924804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,power_law_1.2,0.34895360469818115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,power_law_1.2,0.49015040397644044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,power_law_1.2,0.6494080066680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,power_law_1.01,0.0835968017578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,power_law_1.01,0.10862079858779908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,power_law_1.01,0.18030719757080077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,power_law_1.01,0.2410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,power_law_1.01,0.30550398826599123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,power_law_1.01,0.41417598724365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,power_law_1.01,0.4396160125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,power_law_1.01,0.4409471988677979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,power_law_1.01,0.462662410736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,power_law_1.01,0.47272319793701173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,power_law_1.01,0.48587517738342284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,power_law_1.01,0.493452787399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,power_law_1.01,0.5246143817901612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,power_law_1.01,0.5327680110931396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,power_law_1.01,0.5515391826629639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,power_law_1.01,0.5745855808258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,power_law_1.01,0.6439680099487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,power_law_1.01,0.6506112098693848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,power_law_1.01,0.7548927783966064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,power_law_1.01,0.8029184341430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,power_law_1.01,0.9511551856994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,power_law_1.01,1.0676544189453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,power_law_1.01,1.3735679626464843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,power_law_1.01,1.6664064407348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,power_law_1.01,2.260748863220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,power_law_1.01,2.875545692443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,power_law_1.01,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,power_law_1.01,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,power_law_1.01,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,power_law_1.01,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.01,0.09307519793510437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.01,0.11733759641647339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.01,0.13933440446853637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.01,0.2186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.01,0.2809727907180786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.01,0.43683199882507323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.01,0.5047232151031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,power_law_1.2,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,power_law_1.2,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,power_law_1.2,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,power_law_1.2,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,power_law_1.2,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,power_law_1.2,0.05079039931297302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,power_law_1.2,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,power_law_1.2,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,power_law_1.2,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,power_law_1.2,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,power_law_1.2,0.11125760078430176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,power_law_1.2,0.13102079629898072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,power_law_1.2,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,power_law_1.2,0.3195904016494751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,power_law_1.2,0.4069056034088135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,power_law_1.01,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,power_law_1.01,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,power_law_1.01,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,power_law_1.01,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,power_law_1.01,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,power_law_1.01,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,power_law_1.01,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,power_law_1.01,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,power_law_1.01,0.11693439483642579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.01,0.14732799530029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.01,0.18365440368652344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.01,0.2124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.01,0.28101119995117185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.2,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.2,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.2,0.059596800804138185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.2,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.2,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.2,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.2,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.2,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.2,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.2,0.15044480562210083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.01,0.41320319175720216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.2,0.18407039642333983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.2,0.23452799320220946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.01,0.5251264095306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.2,0.33834879398345946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.2,0.43006720542907717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.01,0.6849792003631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.01,1.1228416442871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.01,1.3515583992004394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,power_law_1.01,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,power_law_1.01,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,power_law_1.01,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,power_law_1.01,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,power_law_1.01,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,power_law_1.01,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,power_law_1.01,0.07856640219688416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,power_law_1.01,0.09441279768943786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,power_law_1.01,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,power_law_1.01,0.13908480405807494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,power_law_1.01,0.16312320232391359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,power_law_1.01,0.2146752119064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,power_law_1.01,0.2791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,power_law_1.01,0.38103039264678956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,power_law_1.01,0.4916031837463379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,balanced,0.05041599770387014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,balanced,0.07285866638024648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,balanced,0.07505066692829132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,balanced,0.07712000111738841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,balanced,0.07512533167997996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,balanced,0.07674666742483775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,balanced,0.07763200004895528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,balanced,0.07915733257929485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,balanced,0.07881600161393483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,balanced,0.08666666348775227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,balanced,0.08685333530108134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,balanced,0.09188266595204671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,balanced,0.09891733527183533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,balanced,0.11078400413195293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,balanced,0.11955733100573222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,balanced,0.16394666830698648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,balanced,0.2188106576601664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,balanced,0.3675413529078166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,balanced,0.46411200364430744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,balanced,0.05286933481693268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,balanced,0.06735999882221222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,balanced,0.08669333656628926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,balanced,0.13548266887664795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,balanced,0.20125865936279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,balanced,0.24958399931589761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,balanced,0.2512320081392924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,balanced,0.25455466906229657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,balanced,0.25407467285792035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,balanced,0.25591466824213666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,balanced,0.25677333275477093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,balanced,0.2593226631482442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,balanced,0.2610293428103129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,balanced,0.26313066482543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,balanced,0.27111466725667316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,balanced,0.2732693354288737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,balanced,0.27802666028340656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,balanced,0.32442132631937665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,balanced,0.3056373397509257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,balanced,0.4289226531982422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,balanced,0.3633333444595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,balanced,0.569546659787496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,balanced,0.46714667479197186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,balanced,0.7638506889343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,balanced,0.6679893334706625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,balanced,0.9730880260467529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,balanced,1.0671306451161702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,balanced,0.027632000545660656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,balanced,0.02749866743882497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,balanced,0.027509334186712902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,balanced,0.0296426663796107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,balanced,0.03203733265399933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,balanced,0.033173332611719765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,balanced,0.032858667274316154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,balanced,0.03518400092919668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,balanced,0.038959999879201256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,balanced,0.03834133346875509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,balanced,0.04138133426507314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,balanced,0.07083733379840851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,balanced,0.09888533751169841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,balanced,0.13693333665529886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,balanced,0.16523733735084534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,balanced,0.09230933586756389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,balanced,0.1246506671110789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,balanced,0.18178667624791464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,balanced,0.2132426699002584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,balanced,0.21815999348958334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,balanced,0.21590399742126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,balanced,0.21529599030812582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,balanced,0.2177706758181254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,balanced,0.22293333212534586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,balanced,0.2272746761639913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,balanced,0.23042666912078857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,balanced,0.23799467086791992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,balanced,0.2479520042737325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,balanced,0.25464532772699994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,balanced,0.26983465751012164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,balanced,0.31276800235112506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,balanced,0.3378346761067708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,balanced,0.4214560190836589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,balanced,0.4754720131556193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,balanced,0.6983679930369059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,balanced,0.8200053373972574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,balanced,1.1825439929962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,balanced,1.517584005991618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,balanced,2.2709973653157554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,balanced,2.856858571370443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,power_law_1.2,0.050355201959609984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,power_law_1.2,0.05065600275993347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,power_law_1.2,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,power_law_1.2,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,power_law_1.2,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.2,0.08590720295906067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.2,0.10925439596176148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.2,0.1235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.2,0.1575103998184204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.2,0.19118720293045044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.2,0.2685823917388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.2,0.36355199813842776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.2,0.48764801025390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.2,0.5923776149749755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,balanced,0.07070933282375336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,balanced,0.11186133821805318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,balanced,0.1634986698627472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,balanced,0.1673333247502645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,balanced,0.16482133666674295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,balanced,0.16341867049535116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,balanced,0.1646453340848287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,balanced,0.16846400499343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,balanced,0.17060800393422446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,balanced,0.17101866006851196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,balanced,0.17575466632843018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,balanced,0.18155733744303384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,balanced,0.18517865737279257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,balanced,0.19366933902104697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,balanced,0.21547732750574747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.2323519984881083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.27404266595840454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.307696004708608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.3943093220392863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.48052799701690674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.7023786703745524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.8834933439890543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,balanced,1.2784106731414795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,balanced,1.6795679728190105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06561920046806335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.09399679899215699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.1410688042640686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.1862912058830261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.2234623908996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,power_law_1.01,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,power_law_1.01,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.31607680320739745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,power_law_1.01,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,power_law_1.01,0.11139199733734131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,power_law_1.01,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,power_law_1.01,0.1736575961112976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,power_law_1.01,0.21495680809020995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,power_law_1.01,0.28935680389404295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,power_law_1.01,0.35978240966796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,power_law_1.01,0.526259183883667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,power_law_1.01,0.6775296211242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.4191103935241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.5986752033233642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.7850815773010253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,power_law_1.2,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,power_law_1.2,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,power_law_1.2,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,power_law_1.2,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,power_law_1.2,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,power_law_1.2,0.04510720074176788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,power_law_1.2,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,power_law_1.2,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,power_law_1.2,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,power_law_1.2,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,power_law_1.2,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,power_law_1.2,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,power_law_1.2,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,power_law_1.2,0.09608319997787476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,power_law_1.2,0.12736639976501465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,power_law_1.2,0.16551040410995482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,power_law_1.2,0.22633600234985352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,power_law_1.2,0.2762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,power_law_1.2,0.4212992191314697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,power_law_1.2,0.5007743835449219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,1.1389375686645509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.4912320137023927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,balanced,0.02739733209212621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,balanced,0.02903466671705246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,balanced,0.029152000943819683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,balanced,0.031898667414983116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,balanced,0.04219200213750204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,balanced,0.05016533533732096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,balanced,0.060138667623202004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,balanced,0.10283199946085612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,balanced,0.14010133345921835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,balanced,0.17955732345581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,balanced,0.24517333507537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,balanced,0.30315200487772626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.1515328049659729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.18001279830932618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.251859188079834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,power_law_1.2,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.32434558868408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,power_law_1.2,0.12614400386810304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,power_law_1.2,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,power_law_1.2,0.12033280134201049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,power_law_1.2,0.11794559955596924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,power_law_1.2,0.1020095944404602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,power_law_1.2,0.11957119703292847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,power_law_1.2,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,power_law_1.2,0.11606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,power_law_1.2,0.11424000263214111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,power_law_1.2,0.1252671957015991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,power_law_1.2,0.11578240394592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,power_law_1.2,0.1150015950202942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,power_law_1.2,0.13169920444488525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,power_law_1.2,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,power_law_1.2,0.1334272027015686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,power_law_1.2,0.1429311990737915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,power_law_1.2,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,power_law_1.2,0.19255679845809937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,power_law_1.2,0.22885758876800538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,power_law_1.2,0.2575040102005005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,power_law_1.2,0.32778880596160886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,power_law_1.2,0.3726912021636963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,power_law_1.2,0.4774144172668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,power_law_1.2,0.6007167816162109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,power_law_1.2,0.8250687599182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,power_law_1.2,1.0844287872314453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,power_law_1.2,0.043263998627662656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,power_law_1.2,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,power_law_1.2,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,power_law_1.2,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,power_law_1.2,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,power_law_1.2,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,power_law_1.2,0.05191680192947388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,power_law_1.2,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,power_law_1.2,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,power_law_1.2,0.1312831997871399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,power_law_1.2,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,power_law_1.2,0.19818880558013915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,power_law_1.2,0.24248321056365968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,power_law_1.2,0.3543040037155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,power_law_1.2,0.4766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.461843204498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.5989312171936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,power_law_1.01,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,power_law_1.01,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,power_law_1.01,0.09045119881629944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,power_law_1.01,0.09958400130271912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,power_law_1.01,0.09967359900474548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,power_law_1.01,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,power_law_1.01,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,power_law_1.01,0.10675840377807617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,power_law_1.01,0.11000319719314575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,power_law_1.01,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,power_law_1.01,0.11896320581436157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,power_law_1.01,0.12322560548782349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,power_law_1.01,0.1363584041595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.01,0.14510079622268676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.01,0.17051520347595214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.01,0.19448319673538209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.01,0.24545280933380126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.01,0.28117120265960693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.01,0.37678720951080324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.01,0.47496957778930665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.01,0.663097620010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.01,0.9117695808410644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,balanced,0.07286400099595387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,balanced,0.07474666833877563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,balanced,0.09167466560999553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,balanced,0.10983467102050781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,balanced,0.14541332920392355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,balanced,0.14637333154678345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,balanced,0.14834133783976236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,balanced,0.1476746698220571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,balanced,0.14998933672904968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,balanced,0.14894933501879373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,balanced,0.15223999818166098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,balanced,0.1530186633268992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,balanced,0.1543786625067393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,balanced,0.1566933294137319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,balanced,0.15845333536465964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,balanced,0.16237866878509521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,balanced,0.16934933265050253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,balanced,0.17966934045155844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,balanced,0.20124799013137817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,balanced,0.21569599707921347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,balanced,0.2521760066350301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,balanced,0.2879679997762044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,balanced,0.4085386594136556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,balanced,0.4916906754175822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,balanced,0.6982506910959879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,balanced,0.9097973505655924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,power_law_1.01,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,power_law_1.01,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,power_law_1.01,0.12880640029907225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,power_law_1.01,0.16807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,power_law_1.01,0.19498239755630492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,power_law_1.01,0.23613440990447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,power_law_1.01,0.2400831937789917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,power_law_1.01,0.2632384061813354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,power_law_1.01,0.2623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,power_law_1.01,0.2688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,power_law_1.01,0.26797440052032473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,power_law_1.01,0.2790656089782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,power_law_1.01,0.2875072002410889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,power_law_1.01,0.3037440061569214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,power_law_1.01,0.3049344062805176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,power_law_1.01,0.3205951929092407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,power_law_1.01,0.3509376049041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,power_law_1.01,0.36419200897216797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,power_law_1.01,0.42229762077331545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,power_law_1.01,0.4468544006347656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,power_law_1.01,0.5530687808990479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,power_law_1.01,0.621779203414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,power_law_1.01,0.8354047775268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,power_law_1.01,0.9723520278930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,power_law_1.01,1.3363840103149414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,power_law_1.01,1.6940031051635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,power_law_1.01,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,power_law_1.01,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,power_law_1.01,0.08828799724578858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,power_law_1.01,0.09240319728851318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,power_law_1.01,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,power_law_1.01,0.10216319561004639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,power_law_1.01,0.10952960252761841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,power_law_1.01,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,power_law_1.01,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,power_law_1.01,0.14307199716567992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,power_law_1.01,0.153164803981781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,power_law_1.01,0.1883263945579529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,power_law_1.01,0.24099841117858886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,power_law_1.01,0.30500481128692625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,power_law_1.01,0.3751424074172974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,power_law_1.01,0.5178880214691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,power_law_1.01,0.6307712078094483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,power_law_1.01,0.9154496192932129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,power_law_1.01,1.1868864059448243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.2,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.2,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.2,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.2,0.04414080083370209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.2,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.2,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.2,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.2,0.1501952052116394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.2,0.18243199586868286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.2,0.2805952072143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.2,0.3037503957748413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,balanced,0.041482667128245033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,balanced,0.041375999649365745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.058975999553998314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07678933441638947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.0972160001595815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.11264533797899882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.1460586686929067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.17917867501576742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,power_law_1.2,0.07747840285301208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,power_law_1.2,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,power_law_1.2,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,power_law_1.2,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,power_law_1.2,0.06919040083885193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,power_law_1.2,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,power_law_1.2,0.07000960111618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,power_law_1.2,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,power_law_1.2,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,power_law_1.2,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,power_law_1.2,0.08583040237426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,power_law_1.2,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,power_law_1.2,0.1093951940536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,power_law_1.2,0.12225919961929321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,power_law_1.2,0.15796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,power_law_1.2,0.1608896017074585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,power_law_1.2,0.21948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,power_law_1.2,0.2710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,power_law_1.2,0.3778559923171997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,power_law_1.2,0.4739391803741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,power_law_1.2,0.7135744094848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,power_law_1.2,0.946662425994873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,balanced,0.048138668139775596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,balanced,0.0925333301226298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,balanced,0.10459733009338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,balanced,0.10769599676132202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,balanced,0.10322667161623637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,balanced,0.10560533404350281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,balanced,0.10589333375295003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,balanced,0.10827199618021648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,balanced,0.11098666985829671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,balanced,0.11361599961916606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,balanced,0.12172266840934753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,balanced,0.12820266683896384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,balanced,0.13873599966367087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,balanced,0.1399786671002706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,balanced,0.16205867131551108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,balanced,0.17693867286046347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,balanced,0.2429813345273336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,balanced,0.25513599316279095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,balanced,0.3792906602223714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,balanced,0.43777068456013996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,balanced,0.6647573312123617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,balanced,0.8002346356709799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,balanced,1.2240586280822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,balanced,1.51255464553833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,power_law_1.01,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,power_law_1.01,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,power_law_1.01,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.01,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.01,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.01,0.10512640476226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.01,0.140665602684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.01,0.1778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.01,0.23441920280456544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.01,0.31025280952453616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.01,0.523686408996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.01,0.6711935997009277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,balanced,0.08345599969228108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,balanced,0.12101333340009053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,balanced,0.18783466021219888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,balanced,0.3333760102589925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,balanced,0.5978453159332275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,balanced,0.8405760129292806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,balanced,0.8438293139139811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,balanced,0.8445440133412679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,balanced,0.8481067021687826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,balanced,0.8540746370951334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,balanced,0.8573599656422933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,balanced,0.8633920351664225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,balanced,0.8678666750590006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,balanced,0.8775520324707031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,balanced,0.8936213652292887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,balanced,0.9010559717814127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,balanced,0.9159253438313802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,balanced,1.049221356709798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,balanced,0.9880586465199789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,balanced,1.374048074086507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,balanced,1.1335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,balanced,1.9279467264811199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,balanced,1.4014132817586262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,balanced,2.061621348063151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,balanced,2.096735954284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,balanced,3.2201334635416665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,balanced,3.4739198684692383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,balanced,0.04613866905371348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,balanced,0.06081599990526835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,balanced,0.06517333288987477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,balanced,0.06689066688219707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,balanced,0.06658133367697398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,balanced,0.07015466690063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,balanced,0.07868800063927968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,balanced,0.0788320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,balanced,0.07956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,balanced,0.10179199775060017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,balanced,0.09754666686058044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,balanced,0.12628799676895142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,balanced,0.15107199549674988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,balanced,0.19429334004720053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,balanced,0.2457866668701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,balanced,0.3444533348083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,balanced,0.45425601800282794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,balanced,0.653989315032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,balanced,0.8556479612986246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,balanced,0.07329600056012471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,balanced,0.09867200255393982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,balanced,0.10006933410962422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,balanced,0.10014933347702026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,balanced,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,balanced,0.10104533036549886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,balanced,0.10386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,balanced,0.1035093367099762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,balanced,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,balanced,0.10756267110506694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,balanced,0.11218133568763733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,balanced,0.11389333009719849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,balanced,0.11932266751925151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,balanced,0.12763733665148416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,balanced,0.13446933031082153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,balanced,0.15265599886576334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,balanced,0.16916799545288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,balanced,0.20289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,balanced,0.23578667640686035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,balanced,0.3094613353411357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,balanced,0.371888001759847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,balanced,0.5159039894739786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,balanced,0.6442133188247681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,power_law_1.01,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,power_law_1.01,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,power_law_1.01,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,power_law_1.01,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,power_law_1.01,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,power_law_1.01,0.07000319957733155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,power_law_1.01,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,power_law_1.01,0.08106240034103393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,power_law_1.01,0.12596479654312134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,power_law_1.01,0.14800000190734863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,power_law_1.01,0.20663039684295653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,power_law_1.01,0.2574912071228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,power_law_1.01,0.3689791917800903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,power_law_1.01,0.47303037643432616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,2,power_law_1.2,0.1146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,2,power_law_1.2,0.125382399559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,power_law_1.2,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,power_law_1.2,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,power_law_1.2,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,power_law_1.2,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,power_law_1.2,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,power_law_1.2,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,power_law_1.2,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,power_law_1.2,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,power_law_1.2,0.0957759976387024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,power_law_1.2,0.1082751989364624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,power_law_1.2,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,power_law_1.2,0.13644800186157227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.2,0.1818112015724182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.2,0.23523199558258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.2,0.2943295955657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.2,0.41083521842956544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.2,0.6216896057128907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.2,0.8188287734985351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.2,1.061683177947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.1520959973335266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.24300799369812012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.4610112190246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.44211840629577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.650278377532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.8054400444030761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,power_law_1.01,0.8004351615905761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,power_law_1.01,0.953337574005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,power_law_1.01,0.9900480270385742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,power_law_1.01,1.0283967971801757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,power_law_1.01,0.9058879852294922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,power_law_1.01,0.9726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,power_law_1.01,0.9597439765930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,power_law_1.01,1.0753664016723632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,power_law_1.01,1.0767168045043944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.0852800369262696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.1117376327514648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.1451904296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.2169343948364257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.3748096466064452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,power_law_1.01,1.4518976211547852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,power_law_1.01,1.6487871170043946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,power_law_1.01,1.8145919799804688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,power_law_1.01,2.272915267944336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,power_law_1.01,2.228371238708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,power_law_1.01,2.8959808349609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,power_law_1.01,3.3525054931640623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.2,1.5455039978027343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,power_law_1.2,0.030188798904418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,power_law_1.2,0.02953599989414215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,power_law_1.2,0.028627198934555054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,power_law_1.2,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,power_law_1.2,0.030534398555755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,power_law_1.2,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,power_law_1.2,0.06366080045700073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,power_law_1.2,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,power_law_1.2,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,power_law_1.2,0.10623999834060668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,power_law_1.2,0.16098560094833375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,power_law_1.2,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.06733440160751343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.06650239825248719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.07562239766120911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.08229759931564332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.09004799723625183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.10387840270996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.12436480522155761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.14663679599761964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.1802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.22724480628967286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.31372799873352053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.3616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.2,2.2169599533081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.5379648208618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,power_law_1.01,0.6576255798339844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,power_law_1.2,0.3449408054351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,power_law_1.2,0.3641599893569946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,power_law_1.01,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,power_law_1.01,0.055622398853302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,power_law_1.01,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,power_law_1.01,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,power_law_1.01,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,power_law_1.01,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,power_law_1.01,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,power_law_1.01,0.16117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,power_law_1.01,0.21509759426116942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,power_law_1.01,0.27508480548858644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,power_law_1.01,0.37122559547424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,power_law_1.01,0.45944957733154296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,power_law_1.01,0.6660799980163574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,power_law_1.01,0.8805439949035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.08833919763565064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.11822719573974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.20238080024719238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.24423680305480958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,balanced,0.029285334050655365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,balanced,0.029365333418051403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2983232021331787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.4151936054229736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.4272064208984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,balanced,0.07029333213965099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,balanced,0.09196266531944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.4285888195037842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,balanced,0.11133333047231038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,balanced,0.15293332934379578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.4725056171417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,balanced,0.19211200873057047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.4899712085723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.478272008895874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.505452823638916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.4973440170288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.5131199836730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.5340928077697754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.5506752014160157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.6021503925323486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.6266367912292481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,0.731609582901001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,0.7912447929382325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,0.9309760093688965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.0833600044250489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.4559616088867187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.6972991943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,2.3023487091064454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,3.1272256851196287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,power_law_1.01,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,power_law_1.01,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,power_law_1.01,0.04211199879646301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,power_law_1.01,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,power_law_1.01,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,power_law_1.01,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,power_law_1.01,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,power_law_1.01,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,power_law_1.01,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,power_law_1.01,0.06547200083732604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,power_law_1.01,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,power_law_1.01,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,power_law_1.01,0.10856959819793702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,power_law_1.01,0.1319551944732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,power_law_1.01,0.18027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,power_law_1.01,0.20588159561157227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,power_law_1.01,0.28126718997955324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,power_law_1.01,0.3481663942337036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,power_law_1.2,0.027020800113677978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,power_law_1.2,0.026310399174690247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,power_law_1.2,0.02794240117073059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,power_law_1.2,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,power_law_1.2,0.028908801078796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,power_law_1.2,0.030195200443267824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,power_law_1.2,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,power_law_1.2,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,power_law_1.2,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,power_law_1.2,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,power_law_1.2,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,power_law_1.2,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,power_law_1.2,0.22337920665740968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,power_law_1.01,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,power_law_1.01,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,power_law_1.01,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,power_law_1.01,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,power_law_1.01,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,power_law_1.01,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,power_law_1.01,0.10746239423751831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,power_law_1.01,0.1168063998222351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,power_law_1.01,0.15263999700546266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,power_law_1.01,0.19395840167999268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,power_law_1.01,0.2532543897628784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,power_law_1.01,0.32395520210266116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,power_law_1.01,0.45325441360473634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,power_law_1.01,0.5709760189056396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.060224002599716185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.11111040115356445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.18769279718399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.21901440620422363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.2900991916656494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.39219839572906495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.5522175788879394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.7311872005462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,power_law_1.2,0.31473920345306394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,power_law_1.2,0.4191423892974854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,balanced,0.0641599992911021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,balanced,0.06746666630109151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,balanced,0.07645333309968312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,balanced,0.07796800136566162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,balanced,0.07469333211580913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,balanced,0.07542400062084198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,balanced,0.07838400204976399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,balanced,0.07670400043328603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,balanced,0.08115733166535695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,balanced,0.08752000331878662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,balanced,0.09274133046468098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,balanced,0.10539199908574422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,balanced,0.11726933717727661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,balanced,0.13616533080736795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,balanced,0.15432533621788025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,balanced,0.19478400548299155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,balanced,0.23035732905069986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,balanced,0.30771734317143756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,balanced,0.378111998240153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,power_law_1.2,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,power_law_1.2,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,power_law_1.2,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,power_law_1.2,0.06270719766616821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,power_law_1.2,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,power_law_1.2,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,power_law_1.2,0.1347648024559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,power_law_1.2,0.16171519756317138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,power_law_1.2,0.24357120990753173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,power_law_1.2,0.27194240093231203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,power_law_1.2,0.49135360717773435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.07100160121917724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.0893567979335785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.08853759765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.09563519954681396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.09063680171966552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.10090880393981934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.10470399856567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.11367679834365844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.11996159553527833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.1407039999961853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.15004160404205322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.17971839904785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.19429759979248046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.2465087890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.29047679901123047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.40611839294433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.5031680107116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.7096511840820312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,0.9299455642700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.07159680128097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,power_law_1.2,0.5928959846496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.09127039909362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.11114239692687988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.13445119857788085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.15310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.2199552059173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.30083839893341063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.44111361503601076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.5839231967926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.8775039672851562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.0942143976688385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.10684159994125367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.19523199796676635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.2525631904602051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.2673151969909668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.290118408203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.2914752006530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.28222079277038575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.2857664108276367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.29492480754852296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.29920639991760256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.3059007883071899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.3189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.33854079246520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.3427135944366455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.37832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.40586237907409667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.4788224220275879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.5414336204528809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,power_law_1.01,0.6825215816497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,power_law_1.01,0.7626431941986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,power_law_1.01,0.957151985168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,power_law_1.01,1.12805118560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,power_law_1.01,1.5310976028442382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,power_law_1.01,1.9155712127685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,balanced,0.05393599967161814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,balanced,0.05490666627883911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,balanced,0.059792002042134605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,balanced,0.06216000020503998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,balanced,0.07010666529337566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,balanced,0.06807466844717662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,balanced,0.07225066423416138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,balanced,0.08109333117802937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,balanced,0.08643733461697896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,balanced,0.09688533345858256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,balanced,0.11492799719174702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,balanced,0.13643200198809305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,balanced,0.15501866738001505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,balanced,0.19951999187469482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,balanced,0.24615466594696045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,balanced,0.3457173506418864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,balanced,0.4215199947357178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,1.1456895828247071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.07083520293235779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.07395200133323669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.10309120416641235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.1394752025604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.16255999803543092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.21042559146881104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.24079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.31928319931030275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.4028287887573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.5776959896087647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.7168575763702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,balanced,0.06532800197601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,balanced,0.0654666672150294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,balanced,0.08814932902654012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,balanced,0.09142933289210002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,balanced,0.09033067027727763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,balanced,0.0916266640027364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,balanced,0.09093333284060161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,balanced,0.09206400314966838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,balanced,0.09390933314959209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,power_law_1.2,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,balanced,0.09563733140627544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,balanced,0.09797333677609761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,balanced,0.11077866951624553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,balanced,0.11272533734639485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,balanced,0.1157919963200887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,balanced,0.133925328652064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,power_law_1.2,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,balanced,0.14018666744232178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,balanced,0.17939200003941855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,power_law_1.2,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,balanced,0.19588265816370645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,balanced,0.264570673306783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,balanced,0.3126773238182068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,balanced,0.44305066267649335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,balanced,0.5732906659444174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,balanced,0.8253493309020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,power_law_1.2,0.06588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,balanced,1.0524426301320393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,power_law_1.2,0.0649728000164032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,power_law_1.2,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,power_law_1.2,0.07672320008277893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,power_law_1.2,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,power_law_1.2,0.08473600149154663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,power_law_1.2,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,power_law_1.2,0.10435839891433715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,power_law_1.2,0.12199679613113404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,power_law_1.2,0.14199680089950562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,power_law_1.2,0.17205120325088502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,power_law_1.2,0.22478079795837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,power_law_1.2,0.2982912063598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,power_law_1.2,0.37518720626831054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,power_law_1.2,0.5270463943481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,power_law_1.2,0.6967360019683838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,power_law_1.2,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,power_law_1.2,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,power_law_1.2,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,power_law_1.2,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,power_law_1.2,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,power_law_1.2,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,power_law_1.2,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,power_law_1.2,0.1415295958518982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,power_law_1.2,0.17214080095291137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,power_law_1.2,0.25647358894348143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,power_law_1.2,0.2971456050872803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,power_law_1.2,0.5825535774230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,power_law_1.2,0.7836544036865234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,balanced,0.03807999938726425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,balanced,0.041989331444104515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,balanced,0.04333333174387614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,balanced,0.0563679983218511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,balanced,0.07053333520889282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,balanced,0.08854933579762776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,balanced,0.11318399508794148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,balanced,0.12974400321642557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,balanced,0.16911466916402182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,balanced,0.21012800931930542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,balanced,0.28884265820185345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,balanced,0.368618647257487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,balanced,0.03108799954255422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,balanced,0.031856000423431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,balanced,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,balanced,0.06263466676076253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,balanced,0.0683840016523997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,balanced,0.06713599960009257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,balanced,0.0710399995247523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,balanced,0.07866666714350383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,balanced,0.07823466757933299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,balanced,0.08301866551240285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,balanced,0.11851200461387634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,balanced,0.1354986627896627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,balanced,0.1742239991823832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,balanced,0.20256000757217407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,balanced,0.28484266996383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,balanced,0.36588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,balanced,0.521232008934021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,balanced,0.6746026674906412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,balanced,0.04820266862710317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,balanced,0.060421332716941833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,balanced,0.09683199723561604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,balanced,0.09649599591890971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,balanced,0.0972106655438741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,balanced,0.09843200445175171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,balanced,0.0993333359559377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,balanced,0.10002133250236511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,balanced,0.09799999992052714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,balanced,0.10400533676147461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,balanced,0.10821866989135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,balanced,0.10951466361681621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,balanced,0.11508799592653911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,balanced,0.12196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.1295199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.162800004084905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.16260799765586853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.20782933632532755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.2300586700439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.32393066088358563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.3727840185165405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.5101173321406046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.6562826633453369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,power_law_1.2,0.02452480047941208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,power_law_1.2,0.025273600220680238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,power_law_1.2,0.024742400646209715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,power_law_1.2,0.027910399436950683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,power_law_1.2,0.02837119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,power_law_1.2,0.029625600576400755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,power_law_1.2,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,power_law_1.2,0.029977598786354066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,power_law_1.2,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,power_law_1.2,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,power_law_1.2,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,power_law_1.2,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,power_law_1.2,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,power_law_1.2,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,power_law_1.2,0.1331455945968628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,power_law_1.2,0.17578879594802857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,power_law_1.2,0.2911871910095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,power_law_1.2,0.35895040035247805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,power_law_1.2,0.5397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.12288639545440674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,balanced,0.07075199981530507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,balanced,0.07985066870848338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,balanced,0.08175999919573466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,balanced,0.08025066554546356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,balanced,0.0814933329820633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,balanced,0.08401599526405334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,balanced,0.08400533596674602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,balanced,0.08979733784993489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,balanced,0.09052800138791402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,balanced,0.09264533718427022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,balanced,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,balanced,0.13005866607030234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,balanced,0.13823999961217245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,balanced,0.1785386602083842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,balanced,0.1653333306312561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,balanced,0.21590399742126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,balanced,0.22738667329152426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,balanced,0.29897065957387287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,balanced,0.344160000483195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.14947839975357055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.19749120473861695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,power_law_1.2,0.7238463878631591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,balanced,0.04836266736189524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,balanced,0.05199466645717621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,balanced,0.09514666597048442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,balanced,0.11627733707427979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,balanced,0.11786133050918579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,balanced,0.12133333086967468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,balanced,0.12220266461372375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,balanced,0.12236799796422322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,balanced,0.12577066818873087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,balanced,0.1317759950955709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,balanced,0.13318399588267008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,balanced,0.1430346667766571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,balanced,0.15055466691652933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.15940266847610474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.17892799774805704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.200655996799469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.23912533124287924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.2900693416595459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.3695840040842692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.4758400122324626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.6375519831975301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.8286506334940592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.22156798839569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.32008960247039797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.47579522132873536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,power_law_1.2,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,power_law_1.2,0.15505919456481934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,power_law_1.2,0.17200640439987183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,power_law_1.2,0.24221439361572267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,power_law_1.2,0.24268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,power_law_1.2,0.24851200580596924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,power_law_1.2,0.2577728033065796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,power_law_1.2,0.26892800331115724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,power_law_1.2,0.2743040084838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,power_law_1.2,0.27365119457244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,power_law_1.2,0.2945728063583374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,power_law_1.2,0.3010495901107788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,power_law_1.2,0.3143232107162476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,power_law_1.2,0.3339200019836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,power_law_1.2,0.37613439559936523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,power_law_1.2,0.37838718891143797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,power_law_1.2,0.4367487907409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,power_law_1.2,0.4714176177978516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,power_law_1.2,0.5858751773834229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,power_law_1.2,0.6573311805725097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,power_law_1.2,0.8512127876281739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,power_law_1.2,1.0160639762878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,power_law_1.2,1.376582431793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,power_law_1.2,1.7578367233276366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,balanced,0.10439999898274739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,balanced,0.11297067006429036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,balanced,0.12873066465059915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,balanced,0.16330666343371072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,balanced,0.23742934068044028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,balanced,0.23881065845489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,balanced,0.24156800905863443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,balanced,0.24250666300455728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,balanced,0.2443093260129293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,balanced,0.24498132864634195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,balanced,0.24813334147135416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,balanced,0.24990934133529663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,balanced,0.25358933210372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,2,balanced,0.2553066611289978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,2,balanced,0.256714661916097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,2,balanced,0.26317866643269855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,2,balanced,0.2774453361829122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,2,balanced,0.28989867369333905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,2,balanced,0.32178666194279987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,2,balanced,0.3507893482844035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,2,balanced,0.41205867131551105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,2,balanced,0.47601068019866943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,2,balanced,0.6680160363515218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,2,balanced,0.796890656153361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,2,balanced,1.130127986272176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,2,balanced,1.4846773147583008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,power_law_1.01,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,power_law_1.01,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,power_law_1.01,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,power_law_1.01,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,power_law_1.01,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,power_law_1.01,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,power_law_1.01,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,power_law_1.01,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,power_law_1.01,0.13204480409622193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,power_law_1.01,0.1433727979660034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,power_law_1.01,0.20062720775604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,power_law_1.01,0.24185600280761718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,power_law_1.01,0.3436608076095581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,power_law_1.01,0.4397119998931885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,power_law_1.01,0.637772798538208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.1479871988296509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.2381824016571045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.34929919242858887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.42916479110717776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.5785280227661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.6873727798461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.7129663944244384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,power_law_1.2,0.7986112117767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,power_law_1.2,0.9539199829101562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,power_law_1.2,0.9012351989746094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,power_law_1.2,0.9734463691711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,power_law_1.2,1.0715328216552735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,power_law_1.2,1.0941247940063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,power_law_1.2,0.9168191909790039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,power_law_1.2,1.0672191619873046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.045132827758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.1767359733581544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.2092543601989747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.2659711837768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,power_law_1.2,1.494099235534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,power_law_1.2,1.530611228942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,power_law_1.2,1.7887039184570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,power_law_1.2,1.9443647384643554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,power_law_1.2,2.3854080200195313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,power_law_1.2,2.4255552291870117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,power_law_1.2,3.130975914001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,power_law_1.2,3.4380863189697264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,power_law_1.01,0.8202048301696777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,power_law_1.01,0.0494271993637085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,power_law_1.01,0.06658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,power_law_1.01,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,power_law_1.01,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,power_law_1.01,0.07932159900665284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,power_law_1.01,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,power_law_1.01,0.08183040022850037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,power_law_1.01,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,power_law_1.01,0.08755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,power_law_1.01,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,power_law_1.01,0.10172159671783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,power_law_1.01,0.10730240345001221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,power_law_1.01,0.12480000257492066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,power_law_1.01,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,power_law_1.01,0.16750080585479737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,power_law_1.01,0.19808640480041503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,power_law_1.01,0.2563071966171265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,power_law_1.01,0.3515968084335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,power_law_1.01,0.4439551830291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,power_law_1.01,0.49277439117431643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,power_law_1.01,0.7131711959838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,power_law_1.01,0.9499327659606933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,power_law_1.01,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,power_law_1.01,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,power_law_1.01,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,power_law_1.01,0.08080000281333924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,power_law_1.01,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,power_law_1.01,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,balanced,0.0965333382288615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,balanced,0.09883200128873189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,balanced,0.09806933005650838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,balanced,0.09908800323804219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,power_law_1.01,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,balanced,0.10107200344403584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,balanced,0.10181333621342976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,balanced,0.10134399930636089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,balanced,0.10299733281135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,balanced,0.10219732920328777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,power_law_1.01,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,balanced,0.10312533378601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,balanced,0.10265066226323445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10739733775456746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,balanced,0.11741866668065389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,balanced,0.11886933445930481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,power_law_1.01,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,balanced,0.11897066235542297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,balanced,0.141594668229421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.14661866426467896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.16992000738779703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,power_law_1.01,0.09884160161018371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.20754667123158774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.265066663424174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.3335946798324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,power_law_1.01,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.4531253178914388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,balanced,0.5664000113805135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,balanced,0.8313759962717692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,power_law_1.01,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,balanced,1.0700746377309163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.01,0.12621439695358277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.01,0.14685440063476562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.01,0.179750394821167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.01,0.21521279811859131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.01,0.27487359046936033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.01,0.3697920083999634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.01,0.49249920845031736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.01,0.6450111865997314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.01,0.9007103919982911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,balanced,0.039690665900707245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,balanced,0.03990933299064636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.05835733314355215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.06667733192443848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.10000532865524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.12408000230789185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.14619732896486917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.19342400630315146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.24194665749867758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,balanced,0.05026666820049286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,balanced,0.05031466484069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,balanced,0.05076266825199127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,balanced,0.05235200126965841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,balanced,0.06041066845258077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,balanced,0.06243200103441874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,balanced,0.06354133288065593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,balanced,0.0703359991312027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,balanced,0.08075200021266937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,balanced,0.08441600203514099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,balanced,0.09126933415730794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,balanced,0.09948800007502238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,balanced,0.10710933804512024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,balanced,0.11966400345166524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,balanced,0.12988266348838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,balanced,0.16411200165748596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,balanced,0.18521066506703696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,balanced,0.23417067527770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,balanced,0.27938665946324664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,balanced,0.3763253291447957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,balanced,0.49006398518880206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.0749504029750824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.07912960052490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.09063680171966552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.09496960043907166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.1252992033958435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.14223999977111818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.14007680416107177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.17745280265808105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.19809919595718384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.25416319370269774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.3183743953704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.42008957862854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.528492784500122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,power_law_1.01,0.753388786315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,power_law_1.01,0.9421631813049316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.2,0.14460159540176393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.2,0.14400639533996581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.2,0.14231040477752685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.2,0.15279359817504884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.2,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.2,0.14436479806900024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.2,0.14673919677734376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.2,0.1434175968170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.2,0.1440000057220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.2,0.14487040042877197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.2,0.15199359655380248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.2,0.1498304009437561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.2,0.15208959579467773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.2,0.1580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.2,0.16874879598617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.2,0.17828480005264283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.2,0.19561599493026732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.2,0.23520638942718505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.2,0.23034238815307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.2,0.2998143911361694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.2,0.33714559078216555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.2,0.4471744060516357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.2,0.4180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.2,0.5047552108764648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.2,0.699283218383789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.2,0.9993087768554687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.2,1.173356819152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.0606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.08259199857711792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.08381440043449402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.07877119779586791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.07767680287361145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.08338559865951538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.08696320056915283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.1752768039703369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.21936640739440919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.25295999050140383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.33596799373626707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.4207808017730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.6159488201141358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.7435840129852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,power_law_1.01,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,power_law_1.01,0.10500479936599731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,power_law_1.01,0.11851520538330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,power_law_1.01,0.1379520058631897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,power_law_1.01,0.1509951949119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,power_law_1.01,0.14683519601821898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,power_law_1.01,0.15630719661712647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,power_law_1.01,0.1577023983001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,power_law_1.01,0.16255999803543092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,power_law_1.01,0.17011200189590453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,power_law_1.01,0.17455359697341918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,power_law_1.01,0.1821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,power_law_1.01,0.1870463967323303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,power_law_1.01,0.19585280418395995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,power_law_1.01,0.21011199951171874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,power_law_1.01,0.22350080013275148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,power_law_1.01,0.25768959522247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,power_law_1.01,0.2818176031112671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,power_law_1.01,0.3497920036315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,power_law_1.01,0.40898561477661133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,power_law_1.01,0.5423488140106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,power_law_1.01,0.6345664024353027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,power_law_1.01,0.8942463874816895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,power_law_1.01,1.120140838623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.2,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.2,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.2,0.1614143967628479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.2,0.19349119663238526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.2,0.2282111883163452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.2,0.21863679885864257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.2,0.23181440830230712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.2,0.21271040439605712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.2,0.22387840747833251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.2,0.22827520370483398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.2,0.24081919193267823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.2,0.24677119255065919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.2,0.24812800884246827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.2,0.2605247974395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.2,0.27146239280700685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.2,0.2754751920700073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.2,0.29456639289855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.2,0.32768640518188474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.2,0.3606911897659302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.2,0.458739185333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.2,0.49439358711242676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.2,0.5704448223114014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.2,0.6563456058502197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.2,0.868876838684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.2,0.9367679595947266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.2,1.404147243499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.01,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.01,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.01,0.0559935986995697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.01,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.01,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.01,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.01,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.01,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.01,0.14602880477905272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.01,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.01,0.20972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.01,0.29516799449920655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.01,0.33994240760803224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,balanced,0.051370665431022644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,balanced,0.05575466652711233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,balanced,0.056474665800730385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,balanced,0.05492799977461497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,balanced,0.06621333460013072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.08509332935015361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.09709333380063374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.12171733379364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.1425333321094513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.18173333009084067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.21625065803527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.29587199290593463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.37349867820739746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.2,1.522105598449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,power_law_1.01,0.027513599395751952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,power_law_1.01,0.025920000672340394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,power_law_1.01,0.026124799251556398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,power_law_1.01,0.026476800441741943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,power_law_1.01,0.027616000175476073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,power_law_1.01,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,power_law_1.01,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,power_law_1.01,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,power_law_1.01,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,power_law_1.01,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,power_law_1.01,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.0745024025440216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.08882560133934021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.1093824028968811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.1382848024368286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.17211519479751586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.2253119945526123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.30227839946746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,balanced,0.0670773337284724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,balanced,0.14246400197347006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,balanced,0.24167466163635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,balanced,0.4405546585718791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,balanced,0.8341333071390787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,balanced,0.8371413548787435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,balanced,0.8351626396179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,balanced,0.8377226988474528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,balanced,0.840453306833903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,balanced,0.8491146564483643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,balanced,0.851365327835083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,balanced,0.8586773077646891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,balanced,0.8623306751251221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,balanced,0.8704799811045328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,balanced,0.8786826928456625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,balanced,0.8939253489176432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,balanced,0.9247626463572184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,balanced,0.9503359794616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,balanced,1.0038879712422688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,balanced,1.0565120379130046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,balanced,1.1651466687520344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,balanced,1.2698826789855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,balanced,1.555173397064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,balanced,1.6565759976704915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,balanced,2.4891626040140786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,balanced,2.7359466552734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.05281280279159546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.052851200103759766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,balanced,0.04013866682847341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,balanced,0.04740799963474274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,balanced,0.07477333148320515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,balanced,0.08113066852092743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,balanced,0.08142933249473572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,balanced,0.08149866759777069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,balanced,0.08144533137480418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,balanced,0.08189866443475087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,balanced,0.08267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,balanced,0.08323733508586884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,balanced,0.08301866551240285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,balanced,0.08328533172607422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,balanced,0.0897706647713979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,balanced,0.0916266640027364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,balanced,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,balanced,0.1181706686814626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,balanced,0.1218773325284322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,balanced,0.15035733580589294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,balanced,0.16275733709335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,balanced,0.22367467482884726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,balanced,0.25262399514516193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,balanced,0.3325600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,balanced,0.4057226578394572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.1309440016746521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.14118399620056152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.18276480436325074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.21994240283966066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.30246400833129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.38034560680389407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,power_law_1.2,0.5445312023162842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,balanced,0.05426133175690969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,balanced,0.07437866429487865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,balanced,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,balanced,0.07840533554553986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,balanced,0.07998933394749959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,balanced,0.08067200084527333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,balanced,0.08106666803359985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,balanced,0.08196266492207845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,balanced,0.08330666522185008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,balanced,0.08556266625722249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,balanced,0.08674133817354839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,balanced,0.0855519970258077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,balanced,0.09075199564297994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,balanced,0.09909333785374959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,balanced,0.09873599807421367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,balanced,0.10315733154614766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,balanced,0.12727466225624084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,balanced,0.13149333000183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,balanced,0.16079466541608176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,balanced,0.18763200441996256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,balanced,0.24196267127990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,balanced,0.2863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,balanced,0.40213334560394287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,balanced,0.5089120070139567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,balanced,0.7376800378163656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,balanced,0.9467946688334147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,power_law_1.2,0.6826111793518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,power_law_1.2,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,power_law_1.2,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,power_law_1.2,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,power_law_1.2,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,power_law_1.2,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,power_law_1.2,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,power_law_1.2,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,power_law_1.2,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,power_law_1.2,0.0991807997226715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,power_law_1.2,0.13296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,balanced,0.05161599814891815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,balanced,0.052906667192777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,balanced,0.10788266857465108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,balanced,0.1986293395360311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,balanced,0.27462400992711383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,balanced,0.34624000390370685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,power_law_1.2,0.20729598999023438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,power_law_1.2,0.23301119804382325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,power_law_1.2,0.302239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.0679423987865448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,power_law_1.2,0.44556798934936526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.06998400092124939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.11816960573196411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.13275519609451295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.16524800062179565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.1962175965309143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.2511039972305298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.33225600719451903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,power_law_1.01,0.4410367965698242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,power_law_1.01,0.5614592075347901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,power_law_1.2,0.6841087818145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,balanced,0.04826666911443075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,balanced,0.05759466687838236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,balanced,0.05658666789531708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,balanced,0.056613331039746605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,balanced,0.058330665032068886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,balanced,0.06164266665776571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,balanced,0.06388266881306966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,balanced,0.0639626681804657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,balanced,0.07217066486676534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,balanced,0.08522133032480876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,balanced,0.11408533652623494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,balanced,0.11235733826955159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,balanced,0.14414933323860168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,balanced,0.16375999649365744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,power_law_1.01,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,power_law_1.01,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,power_law_1.01,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,power_law_1.01,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,power_law_1.01,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,power_law_1.01,0.08634240031242371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,power_law_1.01,0.0867904007434845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,power_law_1.01,0.09938560128211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,power_law_1.01,0.10520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,power_law_1.01,0.12065919637680053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.01,0.14067840576171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.01,0.16855039596557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.01,0.2129215955734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.01,0.27388160228729247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.01,0.3392064094543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.01,0.4769023895263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.01,0.645030403137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.01,0.862713623046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.01,1.2043968200683595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.09734399914741516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.10498559474945068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.15679999589920043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.16639360189437866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.16533119678497316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.1733888030052185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.18048640489578247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.18354560136795045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.18622080087661744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.19576319456100463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.2026815891265869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.2094208002090454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.25873279571533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.28000640869140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.33019518852233887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.3868864059448242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.48876161575317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.6725632190704346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,power_law_1.2,0.799232006072998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,power_law_1.2,0.9263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,power_law_1.2,1.506809616088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,power_law_1.2,1.8791872024536134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,balanced,0.08294400076071422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,balanced,0.08520533641179402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,balanced,0.0953439970811208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,balanced,0.11369066437085469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,balanced,0.139765332142512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,balanced,0.1641813317934672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,balanced,0.16683199008305868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,balanced,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,balanced,0.16633599996566772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,balanced,0.16727999846140543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,balanced,0.1704639991124471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,balanced,0.17289066314697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,balanced,0.17587733268737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,balanced,0.1828320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,balanced,0.18427733580271402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,balanced,0.19155200322469076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,balanced,0.20519999663035074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,balanced,0.22032533089319864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,balanced,0.254202663898468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,balanced,0.2862773338953654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,balanced,0.36615467071533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,balanced,0.43933868408203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,balanced,0.583295981089274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,balanced,0.7780213356018066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,balanced,1.06932266553243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,balanced,1.4099626541137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,power_law_1.01,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,power_law_1.01,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,power_law_1.01,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,power_law_1.01,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,power_law_1.01,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,power_law_1.01,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,power_law_1.01,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.01,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.01,0.16865919828414916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.01,0.21077120304107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.01,0.3071039915084839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.01,0.3972287893295288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.01,0.562879991531372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.01,0.7748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,power_law_1.01,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,power_law_1.01,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,power_law_1.01,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,power_law_1.01,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,power_law_1.01,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,power_law_1.01,0.0850495994091034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,power_law_1.01,0.08936960101127625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,power_law_1.01,0.0929472029209137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,power_law_1.01,0.10785919427871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,power_law_1.01,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,power_law_1.01,0.13824000358581542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,power_law_1.01,0.1609344005584717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,power_law_1.01,0.20581119060516356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,power_law_1.01,0.2543103933334351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,power_law_1.01,0.3394176006317139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,power_law_1.01,0.43834238052368163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,power_law_1.01,0.6355072021484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,power_law_1.01,0.8201855659484864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,balanced,0.07127466797828674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,balanced,0.09731733798980713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,balanced,0.14629333217938742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,balanced,0.24118399620056152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,balanced,0.4412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,balanced,0.6415573358535767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,balanced,0.6350133419036865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,balanced,0.6348160107930502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,balanced,0.6351840098698934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,balanced,0.6385653416315714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,balanced,0.6415626605351766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,balanced,0.6419093211491903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,balanced,0.643392006556193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,balanced,0.6503200133641561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,balanced,0.6526240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,balanced,0.6597866614659628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,balanced,0.6717759768168131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,balanced,0.6817333698272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,balanced,0.7053066889444987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,balanced,0.7308159669240316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,balanced,0.7702773412068685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,balanced,0.8233439922332764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,balanced,0.8996319770812988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,balanced,1.0286826292673747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,balanced,1.162981351216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,balanced,1.5218453407287598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.03932799994945526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.19955840110778808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.3008512020111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.3798079967498779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,balanced,0.05701333284378052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,balanced,0.05782400071620941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.08257600168387096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.10704533259073894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.11358933647473653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.14034133156140646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.16410666704177856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.2165279984474182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.2577226758003235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.35330132643381756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.4330879847208659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,power_law_1.2,0.029216000437736513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,power_law_1.2,0.02876160144805908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,power_law_1.2,0.028441599011421202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,power_law_1.2,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,power_law_1.2,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,power_law_1.2,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,power_law_1.2,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,power_law_1.2,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,power_law_1.2,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,power_law_1.2,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,power_law_1.2,0.13661439418792726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,power_law_1.2,0.1941696047782898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.07159680128097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.1917248010635376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.24618880748748778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.3030208110809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.38034560680389407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,power_law_1.2,0.25700480937957765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.585862398147583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.7400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,power_law_1.2,0.37160959243774416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,power_law_1.2,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,power_law_1.2,0.09129599928855896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,power_law_1.2,0.10769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,power_law_1.2,0.10894080400466918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,power_law_1.2,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,power_law_1.2,0.10842880010604858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,power_law_1.2,0.11239039897918701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,power_law_1.2,0.11853439807891845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,power_law_1.2,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,power_law_1.2,0.12302720546722412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,power_law_1.2,0.12507519721984864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,power_law_1.2,0.13891199827194214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,power_law_1.2,0.1517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,power_law_1.2,0.1777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,power_law_1.2,0.18042240142822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,power_law_1.2,0.23013761043548583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,power_law_1.2,0.24021759033203124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,power_law_1.2,0.31116158962249757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,power_law_1.2,0.4009727954864502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,power_law_1.2,0.5376255989074707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,power_law_1.2,0.7256383895874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,power_law_1.2,0.4464831829071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,power_law_1.2,0.9444607734680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,power_law_1.2,1.2902527809143067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,power_law_1.2,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.01,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,power_law_1.2,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.01,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,power_law_1.2,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.01,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.01,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,power_law_1.2,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,power_law_1.2,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.01,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.01,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.01,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,power_law_1.2,0.07577599883079529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.01,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.01,0.0967296004295349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.01,0.10828800201416015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.01,0.13008639812469483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.01,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.01,0.14400639533996581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.01,0.17857279777526855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,power_law_1.2,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.01,0.22156159877777098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.01,0.2612799882888794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.01,0.3574336051940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.01,0.3973952054977417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,power_law_1.2,0.11324160099029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,power_law_1.2,0.13325439691543578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,balanced,0.042090664307276406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,balanced,0.04877333343029022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,balanced,0.050666665037473045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,balanced,0.06453333298365276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,balanced,0.0727946658929189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,balanced,0.09717333316802979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,balanced,0.11662399768829346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,balanced,0.1580586632092794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,balanced,0.182751993338267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,balanced,0.2510506709416707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,balanced,0.2982293367385864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,balanced,0.4246240059534709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,balanced,0.5339946746826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,balanced,0.09370666742324829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,balanced,0.14270933469136557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,balanced,0.23837866385777792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,balanced,0.4351360003153483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,balanced,0.8268906275431315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,balanced,1.61626132329305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,balanced,2.3999786376953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,balanced,2.393071969350179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,balanced,2.3999306360880532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,balanced,2.4074559211730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,balanced,2.3960746129353843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,balanced,2.4121546745300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,balanced,2.418399969736735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,balanced,2.4310240745544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,balanced,2.445807933807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,balanced,2.4540747006734214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,balanced,2.483882745107015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,balanced,2.5388639767964682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,balanced,2.5668373107910156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,balanced,2.6670347849527993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,balanced,2.734224001566569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,balanced,2.9061972300211587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,balanced,3.059823989868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,balanced,3.45249080657959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,balanced,3.8427947362264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,power_law_1.2,0.21792640686035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,balanced,4.327306747436523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,balanced,5.809125264485677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,balanced,0.051370665431022644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,balanced,0.07036800185839336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,balanced,0.09723732868830363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,balanced,0.1530346671740214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,balanced,0.2608106732368469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,balanced,0.3116746743520101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,balanced,0.3148266673088074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,balanced,0.31701334317525226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,balanced,0.3192960023880005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,balanced,0.31966932614644367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,balanced,0.32446932792663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,balanced,0.3277653257052104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,balanced,0.3285173376401265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,balanced,0.3356320063273112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,balanced,0.3423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,balanced,0.3482666810353597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,balanced,0.3605920076370239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,balanced,0.3856373230616252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,balanced,0.3996106783548991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,balanced,0.4911946853001912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,balanced,0.48873599370320636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,balanced,0.697322686513265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,balanced,0.6894933382670084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,balanced,0.9910720189412435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,balanced,1.157914638519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,balanced,1.6624479293823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,balanced,2.1145760218302407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,power_law_1.2,0.30169599056243895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,balanced,0.03176533430814743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,balanced,0.03515200068553289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,balanced,0.035887998839219414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,balanced,0.07336000104745229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,balanced,0.09092266360918681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,balanced,0.10890666643778484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,balanced,0.15703999996185303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,balanced,0.19317332903544107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,balanced,0.26998400688171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,balanced,0.34565865993499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,power_law_1.2,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,power_law_1.2,0.07671040296554565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,power_law_1.2,0.07686399817466735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,power_law_1.2,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,power_law_1.2,0.08518400192260742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,power_law_1.2,0.08880640268325805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,power_law_1.2,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,power_law_1.2,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,power_law_1.2,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,power_law_1.2,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,power_law_1.2,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,power_law_1.2,0.1329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,power_law_1.2,0.15564160346984862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,power_law_1.2,0.19197440147399902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,power_law_1.2,0.24444160461425782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,power_law_1.2,0.28310399055480956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,power_law_1.2,0.4088895797729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,power_law_1.2,0.516486406326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.2,0.1869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.2,0.22834560871124268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.2,0.23892478942871093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.2,0.24703359603881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.2,0.24483199119567872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.2,0.24778239727020263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.2,0.22648959159851073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.2,0.2321023941040039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.2,0.2244800090789795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.2,0.2241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.2,0.24313600063323976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.2,0.24495999813079833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.2,0.24834558963775635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.2,0.2554368019104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.2,0.26942079067230223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.2,0.2633599996566772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.2,0.28549120426177976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.2,0.3355776071548462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.2,0.34407041072845457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.2,0.42707200050354005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.2,0.5191679954528808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.2,0.6161151885986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.2,0.6082111835479737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.2,0.7590015888214111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.2,1.100038433074951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.2,1.476358413696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.2,1.6408256530761718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,power_law_1.2,0.7258048057556152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,power_law_1.2,0.9071999549865722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.10868480205535888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.11167999505996704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.12640000581741334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.13371520042419432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.16031999588012696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.1852735996246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.2090751886367798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.27778561115264894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.33622400760650634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,power_law_1.01,0.43276801109313967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,power_law_1.01,0.475270414352417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,power_law_1.01,0.6643904209136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,power_law_1.01,0.907686424255371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,power_law_1.2,1.423091220855713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,power_law_1.2,2.023686408996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,power_law_1.2,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,power_law_1.2,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,power_law_1.2,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,power_law_1.2,0.05592319965362549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,power_law_1.2,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.2,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.2,0.12508800029754638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.2,0.14995839595794677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.2,0.1667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.2,0.2378688097000122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.2,0.29706239700317383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.2,0.4013631820678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.2,0.5169087886810303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,balanced,0.08292266726493835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,balanced,0.08593599994977315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,balanced,0.08693333466847737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,balanced,0.08692800005276997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,balanced,0.08824533224105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,balanced,0.08892800410588582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,balanced,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,balanced,0.09179199735323589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,balanced,0.09572266538937886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,balanced,0.09758933385213216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,balanced,0.10239467024803162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,balanced,0.10955199599266052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.11772267023722331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.12971199552218118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.14642666776974997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.17282666762669882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.19952533642450967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.2614453236262004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.3094399968783061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.43405334154764813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.5372906525929769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,power_law_1.01,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,power_law_1.01,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,power_law_1.01,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,power_law_1.01,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,power_law_1.01,0.058937597274780276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,power_law_1.01,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,power_law_1.01,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,power_law_1.01,0.062028801441192626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,power_law_1.01,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,power_law_1.01,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.01,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.01,0.11355520486831665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.01,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.01,0.16944639682769774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.01,0.20508160591125488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.01,0.27905280590057374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.01,0.34951679706573485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.01,0.46462078094482423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.01,0.5925695896148682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.2,0.11752959489822387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.2,0.15255680084228515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.2,0.1819391965866089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.2,0.221068811416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.2,0.28890879154205323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.2,0.3964672088623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.2,0.37854719161987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.2,0.3814784049987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.2,0.39809279441833495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.2,0.402675199508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.2,0.4013887882232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.2,0.42167038917541505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.2,0.4323264122009277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.2,0.44371838569641114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.2,0.4702335834503174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.2,0.4772223949432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.2,0.5135935783386231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.2,0.5641088008880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.2,0.5807360172271728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.2,0.6769152164459229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.2,0.776313591003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,power_law_1.01,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.2,0.9868351936340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,power_law_1.01,0.028992000222206115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,power_law_1.01,0.027859199047088622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,power_law_1.01,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,power_law_1.01,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.2,0.9572159767150878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,power_law_1.01,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.2,1.1855168342590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,power_law_1.01,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,power_law_1.01,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.2,1.5110400199890137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,power_law_1.01,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.2,2.0867584228515623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.2,2.4036479949951173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,power_law_1.01,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,power_law_1.01,0.16381440162658692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,power_law_1.01,0.2170367956161499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,power_law_1.01,0.2714751958847046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,power_law_1.01,0.4063551902770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.07971839904785157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.0808896005153656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.09062399864196777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.11872639656066894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.11592960357666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.1273151993751526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.17996159791946412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.22410879135131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,balanced,0.023269332945346832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,balanced,0.02367999901374181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,balanced,0.023103999594847362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,balanced,0.02372266600529353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,balanced,0.02510933329661687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,balanced,0.025301332275072735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,balanced,0.02531733363866806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,balanced,0.027082666754722595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,balanced,0.0271519993742307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,balanced,0.027514666318893433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,balanced,0.027679999669392902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,power_law_1.01,0.3771071910858154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,balanced,0.034330666065216064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,balanced,0.0403466671705246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,balanced,0.05161066850026449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,balanced,0.09917333722114563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,balanced,0.1241333285967509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,power_law_1.01,0.4087808132171631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,power_law_1.01,0.5823935985565185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,power_law_1.01,0.5209216117858887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,power_law_1.01,0.9128959655761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,power_law_1.01,1.5617152214050294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,balanced,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,balanced,0.05485333502292633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,balanced,0.11098666985829671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,balanced,0.1748853325843811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,balanced,0.3017706672350566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,balanced,0.3031040032704671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,balanced,0.3025546669960022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,balanced,0.303439994653066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,balanced,0.305184006690979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,balanced,0.30504000186920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,balanced,0.31121599674224854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,balanced,0.31031467517217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,balanced,0.31227733691533405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,balanced,0.3202720085779826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,balanced,0.3219573299090068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,balanced,0.32782934109369916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,balanced,0.34233065446217853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,balanced,0.3543359835942586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,balanced,0.38074131806691486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,balanced,0.4115946690241496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,balanced,0.4800373315811157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,balanced,0.5144586563110352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,balanced,0.6471146742502848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,balanced,0.7277599970499674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,balanced,1.009994665781657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,balanced,1.2029173374176025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,power_law_1.01,1.5292863845825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,power_law_1.01,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,power_law_1.01,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,power_law_1.01,0.07240960001945496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,power_law_1.01,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,power_law_1.01,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,power_law_1.01,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,power_law_1.01,0.07684479951858521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,power_law_1.01,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,power_law_1.01,0.08393599987030029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,power_law_1.01,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,power_law_1.01,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.102457594871521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.12780159711837769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.1624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.19701759815216063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.24564480781555176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.29777278900146487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,power_law_1.01,0.41162881851196287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,power_law_1.01,0.5096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,power_law_1.01,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,power_law_1.01,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,power_law_1.01,0.07707520127296448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,power_law_1.01,0.09386879801750184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,power_law_1.01,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,power_law_1.01,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,power_law_1.01,0.10427520275115967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,power_law_1.01,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,power_law_1.01,0.11434880495071412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,power_law_1.01,0.1185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,power_law_1.01,0.12279679775238037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,power_law_1.01,0.12952959537506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,power_law_1.01,0.13678719997406005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,power_law_1.01,0.1474496006965637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,power_law_1.01,0.17258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,power_law_1.01,0.20368640422821044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,power_law_1.01,0.251366400718689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,power_law_1.01,0.29023358821868894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,power_law_1.01,0.37309439182281495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,power_law_1.01,0.5189824104309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,power_law_1.01,0.689631986618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,power_law_1.01,0.9434111595153809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,power_law_1.01,1.3039232254028321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,power_law_1.01,1.6499200820922852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.08863360285758973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.1391487956047058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.1605568051338196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.18477439880371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.25530879497528075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.2582207918167114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.263923192024231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,balanced,0.04642133414745331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,balanced,0.07821333408355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.2730623960494995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,balanced,0.10338667035102844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,balanced,0.10493866602579753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,balanced,0.10546132922172546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,balanced,0.10703999797503154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.2770944118499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.2839807987213135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.2969664096832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.31137919425964355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.33019518852233887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.3360127925872803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,balanced,0.10542933146158855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,balanced,0.10678933064142863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.3486527919769287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,balanced,0.10951466361681621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,balanced,0.11136000355084737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.38903040885925294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,balanced,0.11186666289965312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,balanced,0.11700800061225891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,balanced,0.11849066615104675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,balanced,0.12335466345151265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,balanced,0.13424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.43300480842590333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,balanced,0.14054933190345764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,balanced,0.1764906644821167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,balanced,0.18498132626215616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.4995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,balanced,0.2542826731999715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,balanced,0.23806933561960855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,balanced,0.3075360059738159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.5718207836151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,balanced,0.3566346565882365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,balanced,0.4978080193201701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.7093952178955079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,balanced,0.5735626618067423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.7893311977386475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.043724822998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.3132351875305175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,1.8878656387329102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,2.2985536575317385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.026713600754737853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.04449920058250427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.1118656039237976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.15167360305786132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.17182079553604127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.2699392080307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.18090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.2639616012573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,power_law_1.2,0.08408960103988647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.3523008108139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,power_law_1.2,0.09615359902381897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.5943488121032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6064127922058106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,power_law_1.2,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.6544384002685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,power_law_1.2,0.13872640132904052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.664243221282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,power_law_1.2,0.15159679651260377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.662937593460083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,power_law_1.2,0.15427199602127076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.6856895923614502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,power_law_1.2,0.15800319910049437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,power_law_1.2,0.1607807993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,power_law_1.2,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,power_law_1.2,0.17872639894485473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,power_law_1.2,0.18381439447402953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,power_law_1.2,0.2953792095184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,power_law_1.2,0.20286080837249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,power_law_1.2,0.2097088098526001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.6890048027038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.7154304027557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,power_law_1.2,0.22261760234832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.7389887809753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,power_law_1.2,0.2557568073272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.7668032169342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,power_law_1.2,0.28463358879089357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.7890175819396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,power_law_1.2,0.3439296007156372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.8014911651611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,power_law_1.2,0.4054719924926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,power_law_1.2,0.8600831985473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,power_law_1.2,0.5297279834747315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,power_law_1.2,0.8595199584960938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,power_law_1.2,0.6128896236419678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,power_law_1.2,0.9662655830383301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,power_law_1.2,0.8323264122009277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.0466624259948731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,power_law_1.2,1.0324864387512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,power_law_1.2,1.2511232376098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,power_law_1.2,1.5176511764526368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,power_law_1.2,1.415558433532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,power_law_1.2,1.9765823364257813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,power_law_1.2,1.7533504486083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,power_law_1.2,2.1222208023071287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,power_law_1.2,2.847039985656738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,power_law_1.2,3.591737747192383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,power_law_1.01,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,power_law_1.01,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,power_law_1.01,0.04301440119743347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,power_law_1.01,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,power_law_1.01,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,power_law_1.01,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,power_law_1.01,0.12104959487915039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,power_law_1.01,0.16225919723510743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,power_law_1.01,0.21366400718688966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,power_law_1.01,0.31559679508209226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,power_law_1.01,0.392524790763855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,power_law_1.2,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,power_law_1.2,0.04301440119743347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,power_law_1.2,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,power_law_1.2,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,power_law_1.2,0.05109120011329651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,power_law_1.2,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,power_law_1.2,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,power_law_1.2,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,power_law_1.2,0.09105280041694641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,power_law_1.2,0.12546559572219848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,power_law_1.2,0.15022079944610595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,power_law_1.2,0.2109312057495117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,power_law_1.01,0.10360959768295289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,power_law_1.01,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,power_law_1.01,0.1442304015159607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,power_law_1.01,0.1636288046836853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,power_law_1.01,0.1833407998085022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,power_law_1.01,0.20320000648498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,power_law_1.01,0.23559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,power_law_1.01,0.2438271999359131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,power_law_1.01,0.25002241134643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,power_law_1.2,0.22181119918823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,power_law_1.01,0.2561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,power_law_1.01,0.26289279460906984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,power_law_1.01,0.2680896043777466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,power_law_1.01,0.2723392009735107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,power_law_1.01,0.2806272029876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,power_law_1.01,0.2784575939178467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,power_law_1.01,0.28396799564361574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,power_law_1.01,0.2991744041442871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,power_law_1.01,0.3312191963195801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.01,0.3602495908737183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.01,0.4321152210235596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.01,0.5081664085388183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.01,0.6134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.01,0.7407423973083496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.01,1.00382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.01,1.274073600769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.01,1.811737632751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.01,2.3532032012939452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,power_law_1.2,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,power_law_1.2,0.3989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.03695360124111176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.11292159557342529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,power_law_1.2,0.6410816192626954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.13505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.19879039525985717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.06613119840621948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.26267518997192385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.09158400297164918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.09313920140266418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.09758719801902771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.15656319856643677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.1815168023109436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.21365759372711182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.26494081020355226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.3325952053070068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.45735678672790525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.5935488224029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,power_law_1.2,0.8333375930786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.2041152000427247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.3200704097747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,power_law_1.2,0.7623807907104492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.3960767984390259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,power_law_1.2,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,power_law_1.2,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,power_law_1.2,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,power_law_1.2,0.07333120107650756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,power_law_1.2,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,power_law_1.2,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,power_law_1.2,0.08990079760551453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,power_law_1.2,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,power_law_1.2,0.09192320108413696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,power_law_1.2,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,power_law_1.2,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,power_law_1.2,0.12740479707717894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,power_law_1.2,0.13141759634017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,power_law_1.2,0.17385599613189698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,power_law_1.2,0.18869760036468505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,power_law_1.2,0.24227840900421144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,power_law_1.2,0.3122431993484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,power_law_1.2,0.4659776210784912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,power_law_1.2,0.537286376953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,balanced,0.04728533327579498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,balanced,0.04825599988301595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,balanced,0.054383998115857445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,balanced,0.058687999844551086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,balanced,0.08478933572769165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,balanced,0.10281599561373393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,balanced,0.12988266348838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,balanced,0.1787946621576945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,balanced,0.2201706568400065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,balanced,0.3052906592686971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,balanced,0.3914560079574585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,balanced,0.0804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,balanced,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,balanced,0.08084266881148021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,balanced,0.08124800026416779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,balanced,0.08149866759777069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,balanced,0.08054933448632558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,balanced,0.07989866534868877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,balanced,0.08333866794904073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,balanced,0.0820906658967336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,balanced,0.08878933389981587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,balanced,0.09274666508038838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,balanced,0.09271466732025146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,balanced,0.09583999713261922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,balanced,0.11308800180753072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,balanced,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,balanced,0.15449066956837973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,balanced,0.18541866540908813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,balanced,0.22843732436498007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,balanced,0.3027733365694682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,balanced,0.3706773519515991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,balanced,0.5300639867782593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,balanced,0.7085599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,balanced,0.0573226660490036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,balanced,0.06537599861621857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,balanced,0.06622933348019917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,balanced,0.06604266663392384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,balanced,0.07053333520889282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,balanced,0.07043200234572093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,balanced,0.07430399954319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,balanced,0.09330133597056071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,balanced,0.09687999884287517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,balanced,0.1160533328851064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,balanced,0.1381706694761912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,balanced,0.1933120091756185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,balanced,0.19819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,balanced,0.2600799997647603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,balanced,0.3124106725056966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,balanced,0.42873068650563556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,balanced,0.5324960152308146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,power_law_1.01,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,power_law_1.01,0.04204800128936768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,power_law_1.01,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,power_law_1.01,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,power_law_1.01,0.043219199776649474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,power_law_1.01,0.043263998627662656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,power_law_1.01,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,power_law_1.01,0.04707840085029602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,power_law_1.01,0.050195199251174924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,power_law_1.01,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,power_law_1.01,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,power_law_1.01,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,power_law_1.01,0.10684159994125367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,power_law_1.01,0.12522239685058595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,power_law_1.01,0.17252479791641234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,power_law_1.01,0.19574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,power_law_1.01,0.2632512092590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,power_law_1.01,0.33722240924835206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,balanced,0.03385599950949351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,balanced,0.038032000263532005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,balanced,0.03724266588687897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,balanced,0.03794133414824804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,balanced,0.03937600056330363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,balanced,0.042378668983777366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,balanced,0.04609066744645437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,balanced,0.10062399506568909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,balanced,0.11890666683514912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,power_law_1.2,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,power_law_1.2,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,power_law_1.2,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,power_law_1.2,0.10045440196990967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,power_law_1.2,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,power_law_1.2,0.10446079969406127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,power_law_1.2,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,power_law_1.2,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,power_law_1.2,0.11223039627075196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,power_law_1.2,0.11620479822158813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,power_law_1.2,0.11883519887924195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,power_law_1.2,0.11968640089035035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,power_law_1.2,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,power_law_1.2,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,power_law_1.2,0.18736000061035157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,power_law_1.2,0.21559040546417235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,power_law_1.2,0.29557759761810304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,power_law_1.2,0.36240639686584475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,power_law_1.2,0.5688191890716553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,power_law_1.2,0.7585408210754394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,power_law_1.2,1.1721856117248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,power_law_1.2,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,power_law_1.2,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,power_law_1.2,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,power_law_1.2,0.04139519929885864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,power_law_1.2,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,power_law_1.2,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,power_law_1.2,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,power_law_1.2,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,power_law_1.2,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,power_law_1.2,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,power_law_1.2,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.2,0.0882752001285553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.2,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.2,0.1506943941116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.2,0.20634241104125978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.2,0.2413696050643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.2,0.3816447973251343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,balanced,0.03387733300526937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,balanced,0.035173334181308746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,balanced,0.0345920001467069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,balanced,0.034154665966828666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,balanced,0.03730666637420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,balanced,0.050293331344922386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,balanced,0.052709331115086876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,balanced,0.08302400012811025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,balanced,0.10800533493359883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,power_law_1.2,1.3203712463378907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,balanced,0.12519466876983643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,balanced,0.16854933897654215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,balanced,0.20505066712697348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.2,0.4969600200653076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.2,0.7806975841522217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.2,0.9495871543884278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,balanced,0.04820266862710317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,balanced,0.055061335364977516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,balanced,0.05625066657861074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,balanced,0.06407466530799866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,balanced,0.06629333396752675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,balanced,0.08239466448624928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,balanced,0.0950986643632253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,balanced,0.1461066703001658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,balanced,0.1777013341585795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,balanced,0.24115200837453207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,balanced,0.3022666573524475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,power_law_1.2,0.07668480277061462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,power_law_1.2,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,power_law_1.2,0.08470399975776673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,power_law_1.2,0.09339519739151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,power_law_1.2,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,power_law_1.2,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,power_law_1.2,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,power_law_1.2,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,power_law_1.2,0.10474239587783814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,power_law_1.2,0.11705600023269654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,power_law_1.2,0.14028160572052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,power_law_1.2,0.13871359825134277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,power_law_1.2,0.15861120223999023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,power_law_1.2,0.19326080083847047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,power_law_1.2,0.2682176113128662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,power_law_1.2,0.36769919395446776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,power_law_1.2,0.43340158462524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,power_law_1.2,0.5936768054962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,power_law_1.2,0.6946303844451904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.1632192015647888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.23117439746856688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,power_law_1.2,1.2088064193725585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,power_law_1.2,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,power_law_1.2,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,power_law_1.2,0.0856000006198883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,power_law_1.2,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,power_law_1.2,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,power_law_1.2,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,power_law_1.2,0.09941120147705078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,power_law_1.2,0.10760320425033569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,power_law_1.2,0.1158784031867981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,power_law_1.2,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,power_law_1.2,0.1295040011405945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,power_law_1.2,0.15071359872817994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,power_law_1.2,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,power_law_1.2,0.20849919319152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,power_law_1.2,0.24609920978546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,power_law_1.2,0.3369728088378906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.27394559383392336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,power_law_1.2,0.41893758773803713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,power_law_1.01,0.10973440408706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,power_law_1.01,0.10519039630889893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,power_law_1.01,0.10391039848327636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,power_law_1.01,0.10833920240402221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,power_law_1.01,0.10937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,power_law_1.01,0.11023999452590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,power_law_1.01,0.11148159503936768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,power_law_1.01,0.10898560285568237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,power_law_1.01,0.1149183988571167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,power_law_1.01,0.10940159559249878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,power_law_1.01,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,power_law_1.01,0.12231040000915527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,power_law_1.2,1.4397695541381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,power_law_1.01,0.12649600505828856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,power_law_1.01,0.1306048035621643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,power_law_1.01,0.1432960033416748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.01,0.16628479957580566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.01,0.18288639783859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.01,0.2025279998779297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.01,0.2623552083969116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.01,0.32205440998077395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.01,0.4334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.01,0.562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.01,0.833011245727539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.01,1.0422335624694825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,power_law_1.2,0.5940608024597168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,power_law_1.2,0.7625664234161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,power_law_1.2,1.0922368049621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.3880831956863403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,power_law_1.2,2.3187583923339843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,power_law_1.2,1.4276288032531739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.48956799507141113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,power_law_1.2,2.8823232650756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,power_law_1.2,0.04371199905872345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,power_law_1.2,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,power_law_1.2,0.09352959990501404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,power_law_1.2,0.09894400238990783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,power_law_1.2,0.10387840270996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,power_law_1.2,0.1080896019935608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,power_law_1.2,0.11436159610748291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,power_law_1.2,0.11776000261306763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,power_law_1.2,0.12320640087127685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,power_law_1.2,0.1312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,power_law_1.2,0.14050559997558593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,power_law_1.2,0.14914560317993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,power_law_1.2,0.17390719652175904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,power_law_1.2,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,power_law_1.2,0.25430400371551515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,power_law_1.2,0.2971839904785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,power_law_1.2,0.39416959285736086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,power_law_1.2,0.5309567928314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,power_law_1.2,0.6762112140655517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,power_law_1.2,0.9533120155334472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,power_law_1.2,1.3742079734802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,power_law_1.2,1.718022346496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,power_law_1.2,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,power_law_1.2,0.04530560076236725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,power_law_1.2,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,power_law_1.2,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,power_law_1.2,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,power_law_1.2,0.10021120309829712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,power_law_1.2,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,power_law_1.2,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,power_law_1.2,0.2080319881439209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,power_law_1.2,0.24658560752868652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,power_law_1.2,0.35335679054260255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,power_law_1.2,0.4745344161987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,power_law_1.2,0.6256256103515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,power_law_1.2,0.8361856460571289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.18958719968795776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.2785664081573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.4291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.43683838844299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.4397759914398193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.4341695785522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.446943998336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.45487360954284667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.46495361328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.4788544178009033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.4946303844451904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.5072512149810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.5019264221191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.5432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.5584959983825684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,power_law_1.01,0.6137472152709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,power_law_1.01,0.6927807807922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,power_law_1.01,0.7360767841339111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,power_law_1.01,0.9092032432556152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,power_law_1.01,1.0446080207824706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,power_law_1.01,1.4129535675048828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,power_law_1.01,0.028492799401283263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,power_law_1.01,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,power_law_1.01,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,power_law_1.01,1.6816768646240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,power_law_1.01,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,power_law_1.01,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,power_law_1.01,0.05354239940643311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,power_law_1.01,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,power_law_1.01,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,power_law_1.01,2.385503959655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,power_law_1.01,0.1628864049911499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,power_law_1.01,0.2156991958618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,power_law_1.01,3.1025279998779296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,power_law_1.01,0.282150411605835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,power_law_1.01,0.38648319244384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,power_law_1.01,4.509011077880859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,power_law_1.01,0.47525758743286134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,power_law_1.01,5.957241439819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,power_law_1.01,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,power_law_1.01,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,power_law_1.01,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,power_law_1.01,0.04304639995098114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,power_law_1.01,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,power_law_1.01,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,power_law_1.01,0.06685439944267273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,power_law_1.01,0.07035520076751708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,power_law_1.01,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,power_law_1.01,0.1226304054260254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,power_law_1.01,0.14344320297241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,power_law_1.01,0.19881600141525269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,power_law_1.01,0.215283203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,power_law_1.2,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,power_law_1.2,0.18791680335998534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,power_law_1.2,0.2420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,power_law_1.2,0.345196795463562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,power_law_1.2,0.3618240118026733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,power_law_1.2,0.374182391166687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,power_law_1.2,0.384006404876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,power_law_1.2,0.4062016010284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,power_law_1.2,0.40999040603637693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,power_law_1.2,0.42680959701538085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,power_law_1.2,0.4450687885284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,power_law_1.01,0.3208832025527954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,power_law_1.2,0.4559936046600342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,power_law_1.2,0.47854080200195315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,power_law_1.2,0.5125887870788575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,power_law_1.2,0.5849728107452392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,power_law_1.2,0.6161471843719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,power_law_1.2,0.7277184009552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,power_law_1.2,0.8507391929626464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,power_law_1.2,1.076268768310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,power_law_1.2,1.1942784309387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,power_law_1.2,1.5743295669555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,power_law_1.2,1.941587257385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,power_law_1.2,2.6724031448364256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,power_law_1.2,3.4348094940185545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,power_law_1.01,0.4565631866455078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,power_law_1.01,0.5707839965820313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,power_law_1.01,0.7521664142608643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,power_law_1.01,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,power_law_1.01,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,power_law_1.01,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,power_law_1.01,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,power_law_1.01,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,power_law_1.01,0.04378879964351654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,power_law_1.01,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,power_law_1.01,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,power_law_1.01,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,power_law_1.01,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,power_law_1.01,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,power_law_1.01,0.09898239970207215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,power_law_1.01,0.15322240591049194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,power_law_1.01,0.19370239973068237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,power_law_1.01,0.2719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,power_law_1.01,0.3190399885177612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,power_law_1.01,0.06222079992294312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,power_law_1.01,0.06721280217170715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,power_law_1.01,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,power_law_1.01,0.06881920099258423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,power_law_1.01,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,power_law_1.01,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,power_law_1.01,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,power_law_1.01,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,power_law_1.01,0.10709760189056397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.01,0.11146240234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.01,0.13449599742889404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.01,0.15559040307998656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.01,0.19671679735183717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.01,0.2334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.01,0.32622079849243163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.01,0.4023168087005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.01,0.6243711948394776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.01,0.8227840423583984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,balanced,0.11544533570607503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,balanced,0.1848693291346232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,balanced,0.31616000334421795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,balanced,0.5971733331680298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,balanced,1.118122657140096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,balanced,1.6252533594767253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,balanced,1.6328105926513672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,balanced,1.641642729441325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,balanced,1.6467040379842122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,balanced,1.6568106015523274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,balanced,1.6645065943400066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,balanced,1.6758774121602376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,balanced,1.6887307167053223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,balanced,1.7057226498921711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,balanced,1.7266987164815266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,balanced,1.7450292905171711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,balanced,1.7762667338053386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,balanced,1.9928107261657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,balanced,1.9083786010742188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,balanced,2.343146642049154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,balanced,2.1400853792826333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,balanced,3.5672852198282876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,balanced,2.6444212595621743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,balanced,3.8950932820638022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,balanced,4.001386642456055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,balanced,6.3512318929036455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,balanced,6.784197489420573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,power_law_1.2,0.0710207998752594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,power_law_1.2,0.13319679498672485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,power_law_1.2,0.1326591968536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,power_law_1.2,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,power_law_1.2,0.13898880481719972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,power_law_1.2,0.13668479919433593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,power_law_1.2,0.1421056032180786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,power_law_1.2,0.14522240161895753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,power_law_1.2,0.14291839599609374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,power_law_1.2,0.15100159645080566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,power_law_1.2,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,power_law_1.2,0.1582527995109558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,power_law_1.2,0.1788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,power_law_1.2,0.1801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,power_law_1.2,0.19236479997634887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,power_law_1.2,0.2251136064529419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,power_law_1.2,0.2539007902145386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,power_law_1.2,0.31716480255126955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,power_law_1.2,0.36443519592285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,power_law_1.2,0.48030719757080076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,power_law_1.2,0.5984447956085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.07086079716682434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.08746240139007569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.10104960203170776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.11365760564804077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.13566720485687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.1364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.1749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.19411200284957886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,power_law_1.01,0.25461759567260744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,power_law_1.01,0.3035648107528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,power_law_1.01,0.4688511848449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,power_law_1.01,0.5858240127563477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,power_law_1.2,0.8258879661560059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,power_law_1.2,1.0624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,balanced,0.03528533379236857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,balanced,0.033941333492596946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,balanced,0.03664533297220866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,balanced,0.03598399957021078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,balanced,0.039274667700131737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,balanced,0.043365334471066795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,balanced,0.044213334719340004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.049509331583976746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.0726506660381953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.08826667070388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.09974933664004008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.12743999560674033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.15646933515866598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,power_law_1.2,1.5330240249633789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,power_law_1.01,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,power_law_1.01,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,power_law_1.2,1.9982080459594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,power_law_1.01,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,power_law_1.01,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,power_law_1.01,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,power_law_1.01,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,power_law_1.01,0.10432640314102173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,power_law_1.01,0.15009280443191528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,power_law_1.01,0.2014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,power_law_1.01,0.2271359920501709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,balanced,0.040063999593257904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,balanced,0.03791466603676478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,balanced,0.04112533231576284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,balanced,0.04398933549722036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,balanced,0.061648001273473106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,balanced,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,balanced,0.06087466577688853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,balanced,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,balanced,0.07652799785137177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,balanced,0.08262399832407634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,balanced,0.10635200142860413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,balanced,0.14782399932543436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,balanced,0.16703466574350992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,balanced,0.22941333055496216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,balanced,0.2914453347524007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,balanced,0.43932799498240155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,power_law_1.01,0.40464000701904296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,balanced,0.5766719977060953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,power_law_1.01,0.4402944087982178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.07507200241088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.07813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.07861760258674622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.09187840223312378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.09999359846115112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.11423360109329224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.13420799970626832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.18866560459136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.21950080394744872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.3033344030380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,power_law_1.01,0.3452415943145752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,power_law_1.01,0.5164544105529785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,power_law_1.01,0.659609603881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,balanced,0.05473599831263224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,balanced,0.05641599992911021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,balanced,0.06270933151245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,balanced,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,balanced,0.07660266757011414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,balanced,0.07409599920113881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,balanced,0.0749013324578603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,balanced,0.07617599765459697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,balanced,0.078015998005867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,balanced,0.0825439989566803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,balanced,0.09687466422716777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.10884799559911092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.12356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.13843733072280884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.16869866847991943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.19944000244140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.25618133942286175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.32740267117818195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.44415998458862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.5665173530578613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,power_law_1.01,0.0680512011051178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,power_law_1.01,0.07775359749794006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,power_law_1.01,0.07835519909858704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,power_law_1.01,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,power_law_1.01,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,power_law_1.01,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,power_law_1.01,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,power_law_1.01,0.09021440148353577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,power_law_1.01,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,power_law_1.01,0.11534719467163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.01,0.12531839609146117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.01,0.173363196849823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.01,0.24241280555725098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.01,0.28314239978790284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.01,0.43877758979797366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.01,0.613324785232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.01,0.8663552284240723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.01,1.189510440826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,power_law_1.01,0.029254400730133058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,power_law_1.01,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,power_law_1.01,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,power_law_1.01,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,power_law_1.01,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,power_law_1.01,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,power_law_1.01,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,power_law_1.01,0.09753599762916565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,power_law_1.01,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,power_law_1.01,0.16874239444732667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,power_law_1.01,0.22824959754943847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,power_law_1.01,0.27642240524291994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,power_law_1.01,0.4118015766143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,power_law_1.01,0.5481344223022461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,balanced,0.042175998290379844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,balanced,0.09762133161226909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,balanced,0.10008533795674641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,balanced,0.10342400272687276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,balanced,0.10116799672444661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,balanced,0.10176533460617065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,balanced,0.10290132959683736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,balanced,0.10553066929181416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,balanced,0.1088693340619405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,balanced,0.11096533139546712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,balanced,0.11581333478291829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,balanced,0.12075733145078023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,balanced,0.12588266531626383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,balanced,0.13406399885813394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,balanced,0.15473066767056784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,balanced,0.1674826741218567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,balanced,0.23504000902175903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,balanced,0.258245329062144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,balanced,0.3763680060704549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,balanced,0.47063998381296795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,balanced,0.6872266928354899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,balanced,0.7914453347524008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,balanced,1.1638507048288982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,balanced,1.5145173072814941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.11145600080490112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.15324800014495848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.19804799556732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.23945600986480714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.3072448015213013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.40412158966064454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.5330368041992187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,power_law_1.01,0.7543168067932129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,power_law_1.01,0.8986368179321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.0800383985042572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.09624320268630981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.09701759815216064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.09893760085105896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.13056639432907105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.13772799968719482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.16855679750442504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.20286719799041747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.26794240474700926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.3842047929763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.43378558158874514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,power_law_1.01,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,power_law_1.01,0.07025920152664185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,power_law_1.01,0.09621760249137878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,power_law_1.01,0.09086719751358033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,power_law_1.01,0.10273920297622681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,power_law_1.01,0.11448960304260254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,power_law_1.01,0.12120319604873657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,power_law_1.01,0.1161471962928772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,power_law_1.01,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,power_law_1.01,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,power_law_1.01,0.12848000526428222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,power_law_1.01,0.12726399898529053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,power_law_1.01,0.13348480463027954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,power_law_1.01,0.1390079975128174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,power_law_1.01,0.14010239839553834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,power_law_1.01,0.1548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,power_law_1.01,0.1676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,power_law_1.01,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,power_law_1.01,0.2234368085861206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,power_law_1.01,0.23994879722595214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,power_law_1.01,0.3145024061203003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,power_law_1.01,0.38895359039306643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,power_law_1.01,0.49977598190307615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,power_law_1.01,0.6280896186828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,power_law_1.01,0.892204761505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,power_law_1.01,1.1537728309631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.6519743919372558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.8002431869506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.0908735990524292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.1204416036605835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.13541760444641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.13544960021972657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.14085760116577148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.14594559669494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.17523200511932374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.21980159282684325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.1539520025253296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.16929279565811156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.18367359638214112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.20492799282073976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.2104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.2367487907409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.28370559215545654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,power_law_1.01,0.31889920234680175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,power_law_1.01,0.4528192043304443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.0739071846008301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,power_law_1.01,0.523801612854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,power_law_1.01,0.676038408279419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,power_law_1.01,0.8844032287597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,balanced,0.08271466692288716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,balanced,0.11533332864443462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,balanced,0.14460266629854837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,balanced,0.14633599917093912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,balanced,0.1453120013078054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,balanced,0.1455839971701304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,balanced,0.1486026644706726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,balanced,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,balanced,0.15042666594187418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,balanced,0.15095466375350952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,balanced,0.15928000211715698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,balanced,0.16007999579111734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,balanced,0.16403733690579733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,balanced,0.1740000049273173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.180351992448171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.19902400175730386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.21568532784779867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.2510453263918559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.2876853346824646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.3534613450368245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.4355039993921916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.5673760175704956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.7325493494669596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,power_law_1.01,1.241215991973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,power_law_1.01,1.5235072135925294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.4721088409423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,power_law_1.01,2.5334272384643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,power_law_1.01,3.2660926818847655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,power_law_1.01,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,power_law_1.01,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,power_law_1.01,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,power_law_1.01,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,power_law_1.01,0.07850239872932434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,power_law_1.01,0.09498879909515381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.01,0.11900160312652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.01,0.1354688048362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.01,0.1751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.01,0.20724480152130126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.01,0.28336639404296876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.01,0.38867840766906736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.01,0.5413248062133789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.01,0.6867712020874024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,power_law_1.2,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,power_law_1.2,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,power_law_1.2,0.029996800422668456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,power_law_1.2,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,power_law_1.2,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,power_law_1.2,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,power_law_1.2,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,power_law_1.2,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,power_law_1.2,0.08510079979896545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.08019840121269226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.0974016010761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.08824959993362427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.09208319783210754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.12091519832611083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.1303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.11575679779052735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.15232640504837036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.14394240379333495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,power_law_1.2,0.1449023962020874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,balanced,0.029450667401154835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,balanced,0.02942933390537898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,balanced,0.03014400104681651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,balanced,0.030949334303538006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,balanced,0.03183999905983607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,balanced,0.031930667658646904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,balanced,0.03183466692765554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,balanced,0.03403199960788091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,balanced,0.04237333436806997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,balanced,0.042064001162846885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.1860416054725647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,balanced,0.06451199948787689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,balanced,0.07923733194669087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,balanced,0.1055626670519511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,balanced,0.11945600310961406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,balanced,0.155130664507548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,balanced,0.18745599190394083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.23816959857940673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.2898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,power_law_1.2,0.18036479949951173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,power_law_1.2,0.3428031921386719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,power_law_1.2,0.24866559505462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.14728959798812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.17887359857559204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.25759999752044677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.32814080715179444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,power_law_1.2,0.5154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,power_law_1.2,0.3188544034957886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,power_law_1.2,0.5909632205963135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,power_law_1.2,0.4580543994903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,power_law_1.2,0.8214143753051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,balanced,0.03186666717131933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,balanced,0.03388266762097677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,balanced,0.03417066733042399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,balanced,0.03624533365170161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,balanced,0.03794133414824804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,balanced,0.03813866774241129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,balanced,0.05031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,balanced,0.056618665655454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,balanced,0.06619200110435486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,balanced,0.08082666496435802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,balanced,0.09539199868837993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,balanced,0.12601066629091898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,balanced,0.15589333573977152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,balanced,0.2116373380025228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,balanced,0.254746675491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,balanced,0.3542773326237996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,balanced,0.45402665932973224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.05093119740486145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.09780480265617371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.11706240177154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.13303040266036986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.19454079866409302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.2630399942398071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.32534399032592776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.4625728130340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.6
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,power_law_1.2,0.6077055931091309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,power_law_1.2,1.3941951751708985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,power_law_1.01,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,power_law_1.01,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,power_law_1.01,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,power_law_1.01,0.06581119894981384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,power_law_1.01,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,power_law_1.01,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,power_law_1.01,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,power_law_1.01,0.0826367974281311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,power_law_1.01,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,power_law_1.01,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,power_law_1.01,0.12928639650344848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,power_law_1.01,0.15159679651260377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,power_law_1.01,0.16561280488967894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.01,0.20066559314727783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.01,0.2561471939086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.01,0.35561599731445315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.01,0.5258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,balanced,0.03166933357715607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,balanced,0.028597332537174225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,balanced,0.0310506671667099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,balanced,0.0312266672650973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.01,0.7424384117126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,balanced,0.06861333549022675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,balanced,0.09286399682362874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,balanced,0.12843199570973715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,balanced,0.15844800074895224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,balanced,0.22181334098180136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,balanced,0.26337067286173504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.01,1.0500736236572266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.01,1.2927743911743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.01,1.9754623413085937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,8,balanced,0.12063466509183247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,8,balanced,0.1339893341064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,8,balanced,0.16426133116086325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,8,balanced,0.2424266735712687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,8,balanced,0.3498293161392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,8,balanced,0.4392213424046834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,8,balanced,0.4411253531773885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,8,balanced,0.4405386845270793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,8,balanced,0.4436533451080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,8,balanced,0.4426666498184204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,8,balanced,0.4468746582667033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,8,balanced,0.44945065180460614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,8,balanced,0.44954665501912433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,8,balanced,0.4528426726659139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,8,balanced,0.45791467030843097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,8,balanced,0.4589759906133016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,8,balanced,0.469376007715861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,8,balanced,0.4910506804784139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,8,balanced,0.5075413386027018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,8,balanced,0.5508319934209188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,8,balanced,0.5915040175120035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,8,balanced,0.7848853270212809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,8,balanced,0.8422719637552897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,8,balanced,1.1486293474833171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,8,balanced,1.3071466286977131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,8,balanced,1.94758939743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,8,balanced,2.4308266639709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.01,2.735385513305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,balanced,0.07448000212510426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,balanced,0.17222932974497476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,balanced,0.29950400193532306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,balanced,0.5514613389968872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,balanced,1.0510186354319255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,balanced,1.0534453392028809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,balanced,1.0572693347930908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,balanced,1.0611786842346191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,balanced,1.0625173250834148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,balanced,1.0647786458333333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,balanced,1.0710399945576985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,balanced,1.0715733369191487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,balanced,1.0804479916890461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,balanced,1.091312011082967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,balanced,1.095952033996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,balanced,1.1093546549479167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,balanced,1.1387466589609783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,balanced,1.1620852947235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,balanced,1.2119519710540771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,balanced,1.2882293065388997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,balanced,1.3741973241170247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,balanced,1.4578132629394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,balanced,1.8216053644816081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,balanced,1.8834293683369954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,balanced,2.6043306986490884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,balanced,2.9170878728230796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,power_law_1.2,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,power_law_1.2,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,power_law_1.2,0.029356798529624938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,power_law_1.2,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,power_law_1.2,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,power_law_1.2,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,power_law_1.2,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,power_law_1.2,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,power_law_1.2,0.09983999729156494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,power_law_1.2,0.14814079999923707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,power_law_1.01,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,power_law_1.01,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,power_law_1.01,0.0862335979938507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,power_law_1.01,0.08828160166740417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,power_law_1.01,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,power_law_1.01,0.09831039905548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,power_law_1.01,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,power_law_1.01,0.10298240184783936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,power_law_1.01,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,power_law_1.01,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,power_law_1.01,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,power_law_1.01,0.13679360151290892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,power_law_1.01,0.16183040142059327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,power_law_1.01,0.18172800540924072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,power_law_1.01,0.2304703950881958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,power_law_1.01,0.2604095935821533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,power_law_1.01,0.3706367969512939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.2,0.09959679841995239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.2,0.10577919483184814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,power_law_1.01,0.45447678565979005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.2,0.09811199903488159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.2,0.09795839786529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.2,0.08657919764518737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.2,0.09687680006027222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.2,0.10034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.2,0.09837440252304078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.2,0.10108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.2,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.2,0.10235519409179687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.2,0.10810240507125854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.2,0.10796159505844116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,power_law_1.01,0.5604544162750245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.2,0.10949120521545411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.2,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.2,0.13628799915313722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.2,0.14339840412139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.2,0.17223039865493775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.2,0.172435200214386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.2,0.20570240020751954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.09223039746284485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.2,0.24766080379486083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.09820799827575684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.2,0.33139839172363283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.2,0.33270399570465087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.14566400051116943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.2,0.42344322204589846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.18544000387191772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.2,0.5736576080322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.23040640354156494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.251526403427124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.2,0.7320256233215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.31491200923919677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.32892160415649413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.3367487907409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,power_law_1.01,0.9092800140380859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.34472320079803465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.3628927946090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.3758975982666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.38744320869445803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.3997951984405518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.4085696220397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.4285888195037842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.4874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.5198463916778564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,0.6302015781402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,0.7132351875305176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,0.8975040435791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,power_law_1.2,0.18120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,1.1038975715637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,1.4675711631774901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.2,1.894246482849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.2,2.6015167236328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.2,3.294220733642578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,power_law_1.01,0.9335743904113769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.2,1.0357952117919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,power_law_1.01,1.6822528839111328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,power_law_1.01,2.114035224914551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,power_law_1.2,0.24045441150665284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,balanced,0.04218133290608724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,balanced,0.06039466460545858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,balanced,0.06362133224805196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,balanced,0.061978667974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,balanced,0.06857599814732869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,balanced,0.07443733513355255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,balanced,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,balanced,0.08699733018875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,balanced,0.11329066753387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,balanced,0.14217066764831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,balanced,0.17903467019399008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,balanced,0.2125493288040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,balanced,0.29921066761016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,balanced,0.37433600425720215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,balanced,0.5468426545461019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,balanced,0.696885347366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,balanced,0.03198933353026708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,balanced,0.03586133321126302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,balanced,0.03799466788768768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,balanced,0.044138665000597634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,balanced,0.07175999879837036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,balanced,0.08089600006739299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,balanced,0.10692800084749858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,balanced,0.13181333740552267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,balanced,0.18308266003926596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,balanced,0.2246133287747701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.0573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.06435840129852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.08976640105247498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.12193280458450317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.141702401638031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.21407361030578614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.2979840040206909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.359116792678833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,power_law_1.01,0.5063615798950195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,power_law_1.01,0.6763199806213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,power_law_1.01,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,power_law_1.01,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,power_law_1.01,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,power_law_1.01,0.05976960062980652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,power_law_1.01,0.066975998878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,power_law_1.01,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,power_law_1.01,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,power_law_1.01,0.09943680167198181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,power_law_1.01,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,power_law_1.01,0.16172800064086915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,power_law_1.01,0.21472640037536622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,power_law_1.01,0.2597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,power_law_1.01,0.36484479904174805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,power_law_1.01,0.49818878173828124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,balanced,0.0666240006685257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,balanced,0.0787306676308314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,balanced,0.12174399693806966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,balanced,0.1541920006275177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,balanced,0.15275733669598898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,balanced,0.1548479994138082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,balanced,0.1556533376375834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,balanced,0.1551413337389628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,balanced,0.15678933262825012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,balanced,0.1569493313630422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,balanced,0.15968533356984457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,balanced,0.16102400422096252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,balanced,0.1602720022201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,balanced,0.1644480029741923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,balanced,0.16641599933306375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,balanced,0.1691360076268514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,balanced,0.19824532667795816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,balanced,0.18530666828155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,balanced,0.24243199825286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,balanced,0.22605866193771362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,balanced,0.3407200177510579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,balanced,0.27935999631881714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,balanced,0.3823893467585246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,balanced,0.378661314646403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,balanced,0.5591466824213663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,balanced,0.5829066832860311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,power_law_1.01,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,power_law_1.01,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,power_law_1.01,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,power_law_1.01,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,power_law_1.01,0.05108479857444763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,power_law_1.01,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,power_law_1.01,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,power_law_1.01,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,power_law_1.01,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,power_law_1.01,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.01,0.07223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.01,0.08246399760246277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.01,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.01,0.11686400175094605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.01,0.13979519605636598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.01,0.18876800537109376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.01,0.2275775909423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.01,0.30979840755462645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.01,0.39819519519805907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.06035839915275574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.10609920024871826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.146943998336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.1857856035232544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.24007039070129393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.3101696014404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.0950976014137268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.09827839732170104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.09908480048179627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.09895039796829223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.10094720125198364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.09808639883995056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.11290880441665649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.11989760398864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.13370879888534545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.1456704020500183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.1729151964187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.20167040824890137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.2557375907897949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.3014271974563599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.40843520164489744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.49782400131225585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.6911424160003662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,0.8913087844848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.04966399967670441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.07953280210494995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.0966592013835907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.1081536054611206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.12669440507888793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.15168639421463012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.17445759773254393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.21884799003601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.26409599781036375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.36163840293884275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.4622528076171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6653567790985108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.8633472442626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.261740779876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.6468416213989259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,power_law_1.2,0.03805440068244934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,power_law_1.2,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,power_law_1.2,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,power_law_1.2,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,power_law_1.2,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,power_law_1.2,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,power_law_1.2,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,power_law_1.2,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,power_law_1.2,0.061939197778701785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,power_law_1.2,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.2,0.09338880181312562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.2,0.11862399578094482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.2,0.1381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.2,0.20454399585723876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.2,0.23756160736083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.2,0.39412479400634765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.2,0.5612544059753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.2,0.74235520362854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,balanced,0.04404266675313314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,balanced,0.046480000019073486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,balanced,0.04433600107828776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,balanced,0.049029335379600525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,balanced,0.04761599997679392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,balanced,0.0591893345117569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,balanced,0.058506667613983154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,balanced,0.061792001128196716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,balanced,0.07108800113201141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.08076799909273784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.09196266531944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.10467732946077983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.12407466769218445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.13689600427945456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.1737920045852661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.21121066808700562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.2917226751645406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.380240003267924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.2,1.023033618927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.0531328022480011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.09962239861488342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.11575039625167846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.11842559576034546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.12111999988555908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.1260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.1316480040550232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.13544960021972657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.13661439418792726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.14566400051116943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.15107200145721436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.16269439458847046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.1815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.20857601165771483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.26206719875335693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.3081536054611206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.39343359470367434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.5182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.6840511798858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,power_law_1.01,0.8393024444580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.1553600311279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,power_law_1.01,1.5579327583312987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,power_law_1.01,0.02849920094013214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,power_law_1.01,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,power_law_1.01,0.027961599826812743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,power_law_1.01,0.032051199674606325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,power_law_1.01,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,power_law_1.01,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,power_law_1.01,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,power_law_1.01,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.01,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.01,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.01,0.09226239919662475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.01,0.12924799919128419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.01,0.16543999910354615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.01,0.2414400100708008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.01,0.3163583993911743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.01,0.40671358108520506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.01,0.6712255954742432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.2,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.2,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.2,0.10983040332794189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.2,0.11455359458923339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.2,0.1155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.2,0.12093440294265748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.2,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.2,0.12327680587768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.2,0.12750719785690307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.2,0.1351807951927185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.2,0.13575680255889894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.2,0.14960000514984131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.2,0.1647487998008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.2,0.1619968056678772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.2,0.20164480209350585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.2,0.21788160800933837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.2,0.2624639987945557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.2,0.29441280364990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.2,0.3819648027420044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.2,0.4045695781707764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.2,0.5638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.2,0.7013823986053467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,balanced,0.08258666594823201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,balanced,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,balanced,0.11969600121180217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,balanced,0.14516266187032065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,balanced,0.19894933700561523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,balanced,0.25256532430648804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,balanced,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,balanced,0.09129066268603007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,balanced,0.14189866185188293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,balanced,0.24408000707626343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,balanced,0.4383680025736491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,balanced,0.8267573515574137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,balanced,1.2100000381469727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,balanced,1.209989309310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,balanced,1.212117354075114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,balanced,1.2125439643859863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,balanced,1.2225173314412434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,balanced,1.2287200291951497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,balanced,1.2282400131225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,balanced,1.2318027019500732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,balanced,1.2434293429056804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,balanced,1.2493600050608318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,balanced,1.2549599806467693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,balanced,1.2741119861602783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,balanced,1.2878560225168865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,balanced,1.3222880363464355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,balanced,1.3629226684570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,balanced,1.4282239278157551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,balanced,1.4981120427449544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,balanced,1.6043680508931477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,balanced,1.7932319641113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,balanced,2.0012639363606772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,balanced,2.650368054707845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.01,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.01,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.01,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.01,0.05891839861869812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.01,0.06383360028266907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.01,0.09085440039634704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.01,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.01,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.01,0.1461951971054077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.01,0.1863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.01,0.23622400760650636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.01,0.31933441162109377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.01,0.395961594581604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,balanced,0.042117332418759666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,balanced,0.06053866446018219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,balanced,0.0682666649421056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,balanced,0.08826133608818054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,balanced,0.11367467045783997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,balanced,0.12834133704503378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,balanced,0.1668213407198588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,balanced,0.20600533485412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,balanced,0.2889653245608012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,balanced,0.36535465717315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,balanced,0.034330666065216064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,balanced,0.04050666590531667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,balanced,0.06735999882221222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,balanced,0.07662400106589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,balanced,0.10055466492970784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,balanced,0.1276853382587433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.15656960010528564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.161407995223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.169376003742218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.18384640216827391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.18544000387191772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.18910080194473267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.19503999948501588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.20415360927581788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.2079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.22412800788879395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.2598144054412842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.29041280746459963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3552767992019653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.4079423904418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.5452991962432862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.6757503986358643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.928377628326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.1766016006469726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,1.682694435119629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.1781055450439455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,balanced,0.029338667790095013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,balanced,0.027269333600997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,balanced,0.029861333469549816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,balanced,0.03199466566244761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,balanced,0.0383146678407987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,balanced,0.07692799965540568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,balanced,0.10365333159764607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,balanced,0.1397813359896342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,balanced,0.1691840092341105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,power_law_1.01,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,power_law_1.01,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,power_law_1.01,0.15404160022735597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,power_law_1.01,0.21052160263061523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,power_law_1.01,0.2629568099975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,power_law_1.01,0.3502336025238037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,power_law_1.01,0.36759679317474364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,power_law_1.01,0.3734591960906982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,power_law_1.01,0.38892159461975095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,power_law_1.01,0.4033215999603271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,power_law_1.01,0.4180287837982178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,power_law_1.01,0.42675199508666994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,power_law_1.01,0.4341248035430908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,power_law_1.01,0.457151985168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,power_law_1.01,0.4717696189880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,power_law_1.01,0.5047872066497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,power_law_1.01,0.5619775772094726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,power_law_1.01,0.5852608203887939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,power_law_1.01,0.7004928112030029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,power_law_1.01,0.8142527580261231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,power_law_1.01,1.0430463790893554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,power_law_1.01,1.1816512107849122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,power_law_1.01,1.5530559539794921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,power_law_1.01,1.9217023849487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,power_law_1.01,2.635980796813965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,power_law_1.01,3.3815807342529296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,power_law_1.01,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,power_law_1.01,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,power_law_1.01,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,power_law_1.01,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,power_law_1.01,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,power_law_1.01,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,power_law_1.01,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,power_law_1.01,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,power_law_1.01,0.10793600082397461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,power_law_1.01,0.136518394947052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,power_law_1.01,0.19370880126953124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,power_law_1.01,0.24378879070281984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,balanced,0.048842668533325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,balanced,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,balanced,0.172650674978892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,balanced,0.1748746633529663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,balanced,0.17595734198888144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,balanced,0.17729600270589194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,balanced,0.1763253410657247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,balanced,0.17805866400400797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,balanced,0.17911465962727866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,balanced,0.17780800660451254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,balanced,0.1834026575088501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,balanced,0.18734933932622275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,balanced,0.1917546590169271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,balanced,0.1965386668841044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,balanced,0.2051466703414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.21338667472203574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.2327679991722107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.2553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.2932906746864319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.3346133232116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.4355253378550212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.5050719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.7112106482187907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,balanced,0.8690666357676188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,power_law_1.2,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,power_law_1.2,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,power_law_1.2,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,power_law_1.2,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,power_law_1.2,0.06650239825248719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,power_law_1.2,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,power_law_1.2,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,power_law_1.2,0.07813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,power_law_1.2,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,power_law_1.2,0.08437119722366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,power_law_1.2,0.08787840008735656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,power_law_1.2,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,power_law_1.2,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,power_law_1.2,0.12446080446243286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,power_law_1.2,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,power_law_1.2,0.18709759712219237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,power_law_1.2,0.22910079956054688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,power_law_1.2,0.2938944101333618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,power_law_1.2,0.4057663917541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,power_law_1.2,0.529747200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,power_law_1.2,0.7761663913726806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,power_law_1.2,0.9974783897399903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,power_law_1.01,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,power_law_1.01,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,power_law_1.01,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,power_law_1.01,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,power_law_1.01,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,power_law_1.01,0.08157439827919007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,power_law_1.01,0.08504319787025452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,power_law_1.01,0.08847360014915466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,power_law_1.01,0.09080960154533387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,power_law_1.01,0.09613440036773682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,power_law_1.01,0.10391680002212525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,power_law_1.01,0.11954560279846191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,power_law_1.01,0.12565120458602905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,power_law_1.01,0.16361600160598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,power_law_1.01,0.19325439929962157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,power_law_1.01,0.23781120777130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,power_law_1.01,0.2840192079544067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,power_law_1.01,0.3779903888702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,power_law_1.01,0.44327678680419924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,power_law_1.01,0.6650623798370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,power_law_1.01,0.8382399559020997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,balanced,0.04025600105524063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,balanced,0.05035200218359629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,balanced,0.060319999853769936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.06532266736030579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.10286933183670044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.1286186675230662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.13568533460299173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.1811466614405314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.22777599096298218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.3186826705932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.4065386851628621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.0841152012348175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.09904639720916748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.11335680484771729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.12266240119934083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.13475199937820434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.13654400110244752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.13358720541000366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.13664640188217164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.1439296007156372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.15289599895477296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.1597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.1752575993537903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.19586559534072875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.2134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.2770944118499756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.30398719310760497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.38017280101776124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.441926383972168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.5594751834869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,balanced,0.054458667834599815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,balanced,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,balanced,0.05933333436648051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,balanced,0.06037333110968272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,balanced,0.06679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,balanced,0.07413866619269054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,balanced,0.08767466743787129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,balanced,0.09642666578292847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,balanced,0.1279146671295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,balanced,0.13132799665133157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,balanced,0.16275200247764587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,balanced,0.2015733321507772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,balanced,0.2691520055135091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,balanced,0.33881068229675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,balanced,0.47780267397562665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,balanced,0.6187040011088053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.7721727848052978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.0826623916625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,balanced,0.09154133001963298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,balanced,0.1267039974530538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,balanced,0.13937066992123923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,balanced,0.1411946713924408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,balanced,0.1432266632715861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,balanced,0.14381333192189535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,balanced,0.14486933747927347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,balanced,0.14689600467681885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,balanced,0.14727999766667685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,balanced,0.14855999747912088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,balanced,0.1523360013961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,balanced,0.15577066938082376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,balanced,0.15992533167203268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,balanced,0.16703999042510986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,balanced,0.17773334185282388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.1906613310178121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.25407467285792035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2399946649869283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.31959466139475506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.3498186667760213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.48073601722717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.5829066832860311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.8043733437856039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.0438559850056965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,1.413599967956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.18794239759445192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.15875840187072754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.23435521125793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.2333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.23652479648590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.24771199226379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.28244481086730955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.2967103958129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.32226560115814207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.27305600643157957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.28792319297790525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.2868351936340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.3079168081283569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.31436159610748293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.34990079402923585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.3770431995391846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,power_law_1.01,0.4135424137115479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,power_law_1.01,0.44898557662963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,power_law_1.01,0.6181888103485107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,power_law_1.01,0.6875072002410889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,power_law_1.01,0.8897407531738282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,power_law_1.01,1.2311488151550294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,power_law_1.01,1.7435264587402344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,power_law_1.01,2.4142208099365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,power_law_1.01,3.0536640167236326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,power_law_1.01,4.721728134155273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.16654720306396484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.17052160501480101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.17926399707794188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.16398719549179078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.1723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.16462080478668212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.1807935953140259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.19465600252151488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.18449920415878296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.19516799449920655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.2077120065689087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.22750720977783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.2316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.27436161041259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.2880192041397095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.371232008934021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.10188159942626954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.3894144058227539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.1049280047416687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.1080064058303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,power_law_1.01,0.5030079841613769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,power_law_1.01,0.6135680198669433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.12463999986648559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,power_law_1.01,0.85098237991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.1301375985145569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.1413375973701477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.16631679534912108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,power_law_1.01,1.032096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.19324159622192383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.23331840038299562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.30697600841522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.37624320983886717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.44085121154785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.5775936126708985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,power_law_1.2,0.7960000038146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.188697624206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,power_law_1.2,1.4623616218566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,power_law_1.2,0.10416640043258667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,power_law_1.2,0.11416319608688355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,power_law_1.2,0.13578239679336548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,power_law_1.2,0.1399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,power_law_1.2,0.1427839994430542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,power_law_1.2,0.14371199607849122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,power_law_1.2,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,power_law_1.2,0.15723520517349243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,power_law_1.2,0.16003199815750122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,power_law_1.2,0.16524800062179565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,power_law_1.2,0.17230720520019532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,power_law_1.2,0.1756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,power_law_1.2,0.18729599714279174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,power_law_1.2,0.2075968027114868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.2,0.22593278884887696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.2,0.26781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.15638400316238404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.2,0.30531198978424073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.2,0.3703743934631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.2,0.44094080924987794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.2,0.6118527889251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.2,0.7990784168243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.2,1.1062335968017578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.2,1.4446911811828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.18228479623794555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.22339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.3094655990600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.4405632019042969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,balanced,0.05518933137257894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,balanced,0.05625066657861074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,balanced,0.057114665706952415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,balanced,0.05898666878541311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,balanced,0.061349332332611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,balanced,0.06653333206971486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,balanced,0.07438933352629344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,balanced,0.0728000005086263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,balanced,0.09387200077374776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,balanced,0.09544533491134644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,balanced,0.11702932914098103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,balanced,0.1341973344484965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,balanced,0.16711467504501343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,balanced,0.2025760014851888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,balanced,0.27906133731206256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,balanced,0.34965332349141437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,balanced,0.49475733439127606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,balanced,0.6392533381779989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.584659194946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,2,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,2,power_law_1.2,0.1307968020439148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,2,power_law_1.2,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,2,power_law_1.2,0.183078396320343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,2,power_law_1.2,0.21042559146881104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,2,power_law_1.2,0.25754880905151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,2,power_law_1.2,0.3187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,2,power_law_1.2,0.3161151885986328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,2,power_law_1.2,0.3212928056716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,2,power_law_1.2,0.32706561088562014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,2,power_law_1.2,0.34810240268707277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,2,power_law_1.2,0.3539776086807251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,2,power_law_1.2,0.3652479887008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,2,power_law_1.2,0.36146559715271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,2,power_law_1.2,0.3787264108657837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,2,power_law_1.2,0.3739135980606079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,2,power_law_1.2,0.3933056116104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,2,power_law_1.2,0.4226047992706299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.2,0.44832000732421873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,power_law_1.2,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.2,0.4920896053314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.2,0.5475391864776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,power_law_1.2,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.2,0.6783616065979003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.2,0.7882815837860108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,power_law_1.2,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.2,1.0147839546203614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,power_law_1.2,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.2,1.3272255897521972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,power_law_1.2,0.06090239882469177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.2,1.775551986694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,power_law_1.2,0.07479040026664734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.2,0.08702080249786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.2,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.2,2.173753547668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.2,0.1639232039451599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.2,0.21211519241333007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.2,0.2780159950256348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.2,0.44994559288024905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.2,0.6612224102020263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.2,1.0239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.2,1.254310417175293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,balanced,0.08140799899895985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,balanced,0.08376000324885051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,balanced,0.08411199847857158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,balanced,0.08262399832407634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,balanced,0.08241066833337148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,balanced,0.08292266726493835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,balanced,0.08493333061536153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,balanced,0.0888213316599528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,balanced,0.08796800176302592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,balanced,0.09081066648165385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,balanced,0.09608532985051473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,balanced,0.0960053304831187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,balanced,0.09618666768074036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,balanced,0.10311999917030334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,balanced,0.10134933392206828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,balanced,0.10511466860771179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,balanced,0.12324266632397969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,balanced,0.11953600247701009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,balanced,0.16242133577664694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,balanced,0.1532586713631948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,balanced,0.23121599356333414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,balanced,0.2510293324788411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,balanced,0.35020800431569415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,balanced,0.42428799470265705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,balanced,0.6126293341318766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,balanced,0.8021386464436849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.23686399459838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.43136000633239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.39869439601898193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.4986368179321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.5671679973602295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.5793024063110351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.565388822555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.5175871849060059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.5213183879852294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.5286528110504151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.5753024101257325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.5673984050750732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.5444352149963378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.5485504150390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.5929664134979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.5972864151000976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.6177216053009034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.6966400146484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,power_law_1.01,0.7687935829162598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,power_law_1.01,0.7573376178741456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,power_law_1.01,0.9201472282409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,power_law_1.01,0.8987135887145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,power_law_1.01,1.1047295570373534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,power_law_1.01,1.2769023895263671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,power_law_1.01,1.5899776458740233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,power_law_1.01,1.956153678894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,balanced,0.04127466678619385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,balanced,0.048911998669306435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,balanced,0.04823466638724009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,balanced,0.050527999798456825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,balanced,0.05390933156013489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,balanced,0.058090666929880776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,balanced,0.06809066732724507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,balanced,0.07437866429487865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,balanced,0.08901866277058919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,balanced,0.10507200161616008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,balanced,0.13878400127092996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,balanced,0.1667840083440145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,balanced,0.21990933020909628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,balanced,0.2698826591173808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,power_law_1.2,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,power_law_1.2,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,power_law_1.2,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,power_law_1.2,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,power_law_1.2,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,power_law_1.2,0.06217600107192993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,power_law_1.2,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,power_law_1.2,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,power_law_1.2,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,power_law_1.2,0.13612159490585327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,power_law_1.2,0.1577407956123352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,power_law_1.2,0.2083456039428711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,power_law_1.2,0.2519808053970337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,power_law_1.2,0.37194879055023194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,power_law_1.2,0.4650559902191162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,power_law_1.2,0.6456831932067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,power_law_1.2,0.8534527778625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.14212479591369628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.18517760038375855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.24018559455871583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3336127996444702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.35122559070587156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.36309759616851806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.38276479244232176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.38366079330444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.4252352237701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.4407872200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.45265917778015136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.4678463935852051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.499513578414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.564083194732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.5943039894104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7100543975830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.7976319789886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.0066047668457032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.13885440826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.4439680099487304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.783135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.4157567977905274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.02935676574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,balanced,0.08031466603279114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,balanced,0.12717333436012268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,balanced,0.12654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,balanced,0.12654933333396912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,balanced,0.1267733375231425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,balanced,0.12980799873669943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,balanced,0.12852799892425537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,balanced,0.13424533605575562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,balanced,0.13621333241462708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,balanced,0.13483200470606485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,balanced,0.142384002606074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,balanced,0.14055466651916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,balanced,0.14467199643452963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,balanced,0.15768000483512878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,balanced,0.15901866555213928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,balanced,0.15773333112398782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,balanced,0.20094933112462363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,balanced,0.19985600312550864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,balanced,0.271504004796346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,balanced,0.2793440024058024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,balanced,0.37163734436035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,balanced,0.4611733357111613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,balanced,0.6423093477884928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,balanced,0.8524533112843832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,balanced,1.2113973299662273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,balanced,1.5983360608418782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,power_law_1.2,0.0707647979259491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,power_law_1.2,0.11438720226287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,power_law_1.2,0.1826815962791443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,power_law_1.2,0.19468799829483033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,power_law_1.2,0.19983359575271606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,power_law_1.2,0.20460159778594972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,power_law_1.2,0.20464000701904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,power_law_1.2,0.2124351978302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,power_law_1.2,0.21534080505371095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,power_law_1.2,0.2284224033355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,power_law_1.2,0.23162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,power_law_1.2,0.24549760818481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,power_law_1.2,0.26012799739837644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,power_law_1.2,0.26854400634765624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,power_law_1.2,0.3040064096450806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,power_law_1.2,0.3357952117919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,power_law_1.2,0.41251201629638673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,power_law_1.2,0.47523841857910154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,power_law_1.2,0.6179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,power_law_1.2,0.7572927951812745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,power_law_1.2,1.0295295715332031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,power_law_1.2,1.2945088386535644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,power_law_1.2,1.830303955078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,balanced,0.04663466910521189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,balanced,0.06876266499360402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,balanced,0.06874133149782817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,balanced,0.06916800141334534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,balanced,0.07047999898592631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,balanced,0.07412800192832947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,balanced,0.07806933422883351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,balanced,0.08317866424719493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,balanced,0.08724266290664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,balanced,0.08905599514643352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,balanced,0.10124799609184265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,balanced,0.10590933760007222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,balanced,0.13194666306177774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,balanced,0.14721600214640299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,balanced,0.19513599077860513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,balanced,0.22900799910227457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,balanced,0.3241013288497925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,balanced,0.39019731680552167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,balanced,0.5488266547520956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,balanced,0.7047093709309896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,power_law_1.2,2.2760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,balanced,0.05602666735649109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,balanced,0.06057600180308024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,balanced,0.08461866776148479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,balanced,0.0842080016930898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,balanced,0.08468266328175862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,balanced,0.08080000181992848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,balanced,0.0867680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,balanced,0.0884320040543874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,balanced,0.08921600381533305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,balanced,0.09941333532333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,balanced,0.10115733742713928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,balanced,0.11141866445541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,balanced,0.11924266815185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,balanced,0.13482133547465006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,balanced,0.14670399824778238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,balanced,0.1855413317680359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,balanced,0.21425066391626993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,balanced,0.2721760074297587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,balanced,0.33632532755533856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,balanced,0.4515093167622884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,balanced,0.5846826632817587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,balanced,0.027424000203609467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,balanced,0.02754666656255722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,balanced,0.03263466556866964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,balanced,0.032357332607110344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,balanced,0.03932799895604452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,balanced,0.042175998290379844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,balanced,0.044250667095184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,balanced,0.06282666822274525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,balanced,0.10522133111953735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,balanced,0.14074132839838663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,balanced,0.17502933740615845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,balanced,0.044853334625562034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,balanced,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,balanced,0.06718400120735168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,balanced,0.0666720022757848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,balanced,0.06858666737874348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,balanced,0.0728053351243337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,balanced,0.07627733548482259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,balanced,0.08053866525491078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,balanced,0.08426666259765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,balanced,0.0858026643594106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,balanced,0.10883200168609619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,balanced,0.11643200119336446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,balanced,0.15034133195877075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,balanced,0.1869279940923055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,balanced,0.2786613305409749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,balanced,0.36075735092163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,balanced,0.5088479916254679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,balanced,0.660261352856954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,balanced,0.9532693227132162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,balanced,1.2411200205485027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,balanced,0.02980799973011017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,balanced,0.030479999879995983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,balanced,0.04061333338419596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,balanced,0.041365332901477814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,balanced,0.050581331054369606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,balanced,0.05645333230495453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,balanced,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,balanced,0.12436266740163167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,balanced,0.15376533071200052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,balanced,0.19580266873041788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,balanced,0.24456000328063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,balanced,0.34414398670196533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,balanced,0.43623467286427814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,balanced,0.06250133117039998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,balanced,0.09106133381525676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,balanced,0.12925333778063455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,balanced,0.19977599382400513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,balanced,0.20345600446065268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,balanced,0.19955732425053915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,balanced,0.20204265912373862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,balanced,0.2031573255856832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,balanced,0.2104746699333191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,balanced,0.21799467007319132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,balanced,0.21693867444992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,balanced,0.22077866395314535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,balanced,0.22553066412607828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,balanced,0.2302666703859965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,balanced,0.2448373238245646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,balanced,0.27026132742563885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,balanced,0.29090134302775067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,balanced,0.34563199679056805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,balanced,0.3959253231684367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,balanced,0.5318666696548462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,balanced,0.609994649887085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,balanced,0.8601546287536621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,balanced,1.0565760135650635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,balanced,1.5156906445821126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,balanced,1.9517439206441243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,balanced,0.06640000144640605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,balanced,0.07308800021807353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,balanced,0.07496533294518788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,balanced,0.07546666761239369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,balanced,0.07561066746711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,balanced,0.07603733241558075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,balanced,0.07868266602357228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,balanced,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,balanced,0.08074666559696198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,balanced,0.08287466565767924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,balanced,0.0888426701227824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,balanced,0.09142399827639262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.0960693359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.11146666606267293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.1176533301671346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.14459733168284097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.15657599767049155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.20233599344889322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.2311733365058899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.3283039927482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.39857598145802814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,balanced,0.03389333436886469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,balanced,0.03306133300065994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,balanced,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,balanced,0.04082666585842768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,balanced,0.04673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,balanced,0.059994667768478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,balanced,0.06607999900976817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,balanced,0.08805867036183675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,balanced,0.10873599847157796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,balanced,0.05522133409976959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,balanced,0.13747732837994894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,balanced,0.13381333152453104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,balanced,0.13453867038091025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,balanced,0.13776000340779623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,balanced,0.13725333412488303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,balanced,0.14224533240000406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,balanced,0.1399626632531484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,balanced,0.14431466658910116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,balanced,0.14904533823331198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,balanced,0.1556106706460317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,balanced,0.16149333119392395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,balanced,0.167738676071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,balanced,0.1806346575419108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,balanced,0.20305599768956503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,balanced,0.21626667181650797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,balanced,0.2804693380991618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,balanced,0.3014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,balanced,0.4183093309402466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,balanced,0.4963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,balanced,0.680357297261556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,balanced,0.8721386591593424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,balanced,1.2656853199005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,balanced,1.623194694519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.08803200125694274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.15887999534606934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.22729599475860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.2869951963424683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,power_law_1.01,0.39096319675445557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,power_law_1.01,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,power_law_1.01,0.11134079694747925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,power_law_1.01,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,power_law_1.01,0.11754879951477051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,power_law_1.01,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,power_law_1.01,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,power_law_1.01,0.11461759805679321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,power_law_1.01,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,power_law_1.01,0.11627520322799682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,power_law_1.01,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,power_law_1.01,0.13326079845428468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,power_law_1.01,0.13756799697875977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,power_law_1.01,0.14154239892959594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,power_law_1.01,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.01,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.01,0.1868607997894287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.01,0.2099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.01,0.2721856117248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.01,0.348140811920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.01,0.4514624118804932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.01,0.6121856212615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.01,0.8188223838806152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.01,1.076480007171631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.08291839957237243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.08872320055961609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.11456639766693115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.15392639636993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.20295679569244385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,power_law_1.01,0.5132927894592285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.24997758865356445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,power_law_1.2,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,power_law_1.2,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,power_law_1.2,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,power_law_1.2,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,power_law_1.2,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,power_law_1.2,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,power_law_1.2,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,power_law_1.2,0.08509439826011658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,power_law_1.2,0.08954240083694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,power_law_1.2,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.13024640083312988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.13634560108184815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.1730944037437439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.21454720497131347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.35498239994049074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.2715840101242065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.3342335939407349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,power_law_1.2,0.4329216003417969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.48559999465942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,power_law_1.2,0.5531583786010742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,balanced,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,balanced,0.08162133395671844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,balanced,0.08334933718045552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,balanced,0.08205866813659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,balanced,0.08642666538556416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,balanced,0.08747733632723491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,balanced,0.08709866801897685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,balanced,0.09493333101272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,balanced,0.09658666451772054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,balanced,0.1072213351726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,balanced,0.11078932881355286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,balanced,0.11993599931399028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,balanced,0.14095466335614523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,balanced,0.15987199544906616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,balanced,0.20048532883326212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,balanced,0.24032533168792725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,balanced,0.31886400779088336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,balanced,0.40908265113830566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,balanced,0.5650080045064291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,balanced,0.7216959794362386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,balanced,0.19119999806086221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,balanced,0.3213919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,balanced,0.6095786492029825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,balanced,1.1259573300679524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,balanced,1.614474614461263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,balanced,1.6146400769551594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,balanced,1.6209759712219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,balanced,1.624351978302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,balanced,1.6307360331217449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,balanced,1.636917273203532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,balanced,1.643466631571452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,balanced,1.652224063873291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,balanced,1.6669546763102214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,balanced,1.6815786361694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,balanced,1.6969173749287922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,balanced,1.7166773478190105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,balanced,1.9063733418782551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,balanced,1.805888017018636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,balanced,2.218272050221761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,balanced,1.9912586212158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,balanced,3.0566720962524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,balanced,2.3264586130777993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,balanced,3.8640000025431314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,balanced,3.4583571751912436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,balanced,5.962970733642578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,balanced,5.273514747619629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,power_law_1.01,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,power_law_1.01,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,power_law_1.01,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,power_law_1.01,0.030534398555755616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,power_law_1.01,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,power_law_1.01,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,power_law_1.01,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,power_law_1.01,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,power_law_1.01,0.0967296004295349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,power_law_1.01,0.12769919633865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,power_law_1.01,0.18456319570541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,power_law_1.01,0.202675199508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,power_law_1.01,0.30901761054992677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,power_law_1.01,0.3930367946624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,balanced,0.061850666999816895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,balanced,0.06359999875227611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,balanced,0.066021333138148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,balanced,0.0658186674118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,balanced,0.07272533575693767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,balanced,0.09070400396982829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,balanced,0.09275199969609578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,balanced,0.11373333136240642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,balanced,0.12352533141771953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,balanced,0.15627200404802957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,balanced,0.18482667207717896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,balanced,0.25358933210372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,balanced,0.2892693281173706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,balanced,0.411626656850179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,balanced,0.5065493186314901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,power_law_1.01,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,power_law_1.01,0.1075711965560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,power_law_1.01,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,power_law_1.01,0.2359679937362671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,power_law_1.01,0.3176703929901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,power_law_1.01,0.368556809425354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,power_law_1.01,0.47013120651245116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,power_law_1.01,0.4820096015930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,power_law_1.01,0.4950719833374023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,power_law_1.01,0.5166783809661866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,power_law_1.01,0.5366655826568604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,power_law_1.01,0.559500789642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,power_law_1.01,0.5698815822601319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,power_law_1.01,0.5905727863311767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,power_law_1.01,0.6091775894165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,power_law_1.01,0.6456704139709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,power_law_1.01,0.6998655796051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,power_law_1.01,0.7673088073730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,power_law_1.01,0.9008511543273926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,power_law_1.01,0.9787712097167969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,power_law_1.01,1.1981568336486816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,power_law_1.01,1.427462387084961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,power_law_1.01,1.868217658996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,power_law_1.01,2.2213632583618166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,power_law_1.01,2.9663040161132814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,power_law_1.01,3.865427017211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,balanced,0.07075199981530507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,balanced,0.09782933195432027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,balanced,0.1369439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,balanced,0.14018133282661438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,balanced,0.14249066511789957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,balanced,0.14260799686113992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,balanced,0.1416853368282318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,balanced,0.14404267072677612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,balanced,0.14615999658902487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,balanced,0.14664533734321594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,balanced,0.15452800194422403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,balanced,0.15942399700482687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,balanced,0.16119466225306192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,balanced,0.1629813313484192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,balanced,0.1737920045852661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,balanced,0.18713066975275675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,balanced,0.20865066846211752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,balanced,0.23035200436909994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,balanced,0.268885334332784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,balanced,0.31307733058929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,balanced,0.420687993367513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,balanced,0.4885813395182292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,balanced,0.6936586697896322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,balanced,0.8634133338928223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.07947520017623902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.0821183979511261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.122707200050354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.12934399843215943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.1311743974685669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.13285759687423707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.13936640024185182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.13948800563812255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.1459391951560974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.14647680521011353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.15103360414505004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.16749440431594848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.18823039531707764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.2292992115020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.2732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,2,power_law_1.01,0.10741759538650512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.3668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.4665215969085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,2,power_law_1.01,0.13148800134658814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,2,power_law_1.01,0.15000319480895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.651635217666626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,2,power_law_1.01,0.18775039911270142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.8263615608215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,2,power_law_1.01,0.21953918933868408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,1.2241536140441895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,2,power_law_1.01,0.25369598865509035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,1.6035520553588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,2,power_law_1.01,0.328217601776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,2,power_law_1.01,0.3339967966079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,2,power_law_1.01,0.3364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,2,power_law_1.01,0.3404927968978882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,2,power_law_1.01,0.3444416046142578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,2,power_law_1.01,0.3539520025253296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,2,power_law_1.01,0.35715200901031496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,2,power_law_1.01,0.37751040458679197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,2,power_law_1.01,0.3698751926422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,2,power_law_1.01,0.3599488019943237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,2,power_law_1.01,0.38061439990997314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,2,power_law_1.01,0.41096320152282717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.01,0.4287871837615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.01,0.47020158767700193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.01,0.516812801361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.01,0.6303487777709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.01,0.7341695785522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.01,1.001689624786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.01,1.1765119552612304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.01,1.7190847396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.01,2.111244773864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,power_law_1.2,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,power_law_1.2,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,power_law_1.2,0.06481279730796814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,power_law_1.2,0.07928959727287292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,power_law_1.2,0.08481919765472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,power_law_1.2,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,power_law_1.2,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,power_law_1.2,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,power_law_1.2,0.0933184027671814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,power_law_1.2,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,power_law_1.2,0.10359040498733521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,power_law_1.2,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,power_law_1.2,0.12283519506454468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,power_law_1.2,0.13775999546051027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,power_law_1.2,0.15838079452514647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,power_law_1.2,0.18318079710006713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,power_law_1.2,0.21287040710449218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,power_law_1.2,0.25657598972320556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,power_law_1.2,0.34571518898010256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,power_law_1.2,0.4834303855895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,power_law_1.2,0.6000832080841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,power_law_1.2,0.8433983802795411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,power_law_1.2,1.0587008476257325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,balanced,0.04117333392302195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,balanced,0.049498667319615684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,balanced,0.05053333441416422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.09712533156077068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.1113866666952769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.13704533378283182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.16430933276812235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.22125866015752158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.27583465973536175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,balanced,0.05842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,balanced,0.09276266892751057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,balanced,0.13758933544158936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,balanced,0.2109066645304362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,balanced,0.28193066517512005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,balanced,0.2834986646970113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,balanced,0.28757866223653156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,balanced,0.2914133270581563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,balanced,0.2916160027186076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,balanced,0.3008693257967631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,balanced,0.3022986650466919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,balanced,0.30878933270772296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,balanced,0.31441599130630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,balanced,0.3197653293609619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,balanced,0.3341226577758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,balanced,0.3780906597773234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,balanced,0.40411198139190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,balanced,0.4706079959869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,balanced,0.5314666827519735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,balanced,0.6668907006581625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,balanced,0.8637493451436361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,balanced,1.1517493724822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,balanced,1.4949439366658528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,balanced,2.1154026985168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,balanced,2.823450724283854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,power_law_1.2,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,power_law_1.2,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,power_law_1.2,0.10554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,power_law_1.2,0.13041919469833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,power_law_1.2,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,power_law_1.2,0.14053759574890137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,power_law_1.2,0.15384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,power_law_1.2,0.1475648045539856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,power_law_1.2,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,power_law_1.2,0.1626431941986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,power_law_1.2,0.15020159482955933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,power_law_1.2,0.1598912000656128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,power_law_1.2,0.16660480499267577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,power_law_1.2,0.16640000343322753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,power_law_1.2,0.17825280427932738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,power_law_1.2,0.18014719486236572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,power_law_1.2,0.19100159406661987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,power_law_1.2,0.2105407953262329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,power_law_1.2,0.21559040546417235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,power_law_1.2,0.25703039169311526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,power_law_1.2,0.27105278968811036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,power_law_1.2,0.3232959985733032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,power_law_1.2,0.35016961097717286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,power_law_1.2,0.4749887943267822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,power_law_1.2,0.5371327877044678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,power_law_1.2,0.6901504039764405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,power_law_1.2,0.8503680229187012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,power_law_1.2,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,power_law_1.2,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,power_law_1.2,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,power_law_1.2,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,power_law_1.2,0.07774720191955567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,power_law_1.2,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,power_law_1.2,0.0917952001094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,power_law_1.2,0.1046336054801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,power_law_1.2,0.11366399526596069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,power_law_1.2,0.12636799812316896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,power_law_1.2,0.139520001411438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,power_law_1.2,0.1702015995979309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,power_law_1.2,0.18628480434417724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,power_law_1.2,0.24661118984222413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,power_law_1.2,0.31256320476531985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,power_law_1.2,0.4473855972290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,power_law_1.2,0.5704832077026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,power_law_1.2,0.8637503623962403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.07731840014457703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.10230400562286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.10987520217895508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,balanced,0.06896000107129414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,balanced,0.1058026651541392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,balanced,0.15434666474660239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,balanced,0.17926400899887085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,balanced,0.18032532930374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,balanced,0.1822133262952169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,balanced,0.18308266003926596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,balanced,0.18102399508158365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,balanced,0.18492799997329712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,balanced,0.18707199891408285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.1075711965560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,balanced,0.18737600247065225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,balanced,0.18752533197402954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,balanced,0.19612266620000204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,balanced,0.1965493361155192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,balanced,0.20458666483561197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,balanced,0.22613867123921713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,balanced,0.22507200638453165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,balanced,0.2624373237291972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,balanced,0.2720106641451518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.129312002658844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,balanced,0.3448266585667928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,balanced,0.3644426663716634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,balanced,0.4907626708348592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,balanced,0.5810559988021851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,balanced,0.7677760124206543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,balanced,1.001312017440796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.13308800458908082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.16505600214004518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,power_law_1.2,1.0366463661193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.06592640280723572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.14958080053329467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.1990399956703186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.2376960039138794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,power_law_1.01,0.30846080780029295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,power_law_1.01,0.4202688217163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.24513919353485109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.2881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,power_law_1.2,0.3991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,power_law_1.2,0.4984447956085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,power_law_1.2,0.7071040153503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,power_law_1.2,0.8972543716430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,power_law_1.01,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,power_law_1.01,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,power_law_1.01,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,power_law_1.01,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,power_law_1.01,0.15274879932403565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,power_law_1.01,0.1826815962791443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,power_law_1.01,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,power_law_1.01,0.19444479942321777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,power_law_1.01,0.2026815891265869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,power_law_1.01,0.21600639820098877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,power_law_1.01,0.22595200538635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,power_law_1.01,0.2436352014541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,power_law_1.01,0.26487040519714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,power_law_1.01,0.28008959293365476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,power_law_1.01,0.3104896068572998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,power_law_1.01,0.3864768028259277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,power_law_1.01,0.3697472095489502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,power_law_1.01,0.47132158279418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,power_law_1.01,0.5497024059295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,power_law_1.01,0.7338880062103271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,power_law_1.01,0.9143424034118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,power_law_1.01,1.248416042327881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,power_law_1.01,1.5734399795532226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,power_law_1.01,2.2490047454833983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,power_law_1.01,2.8910591125488283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,power_law_1.2,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,power_law_1.2,0.04414080083370209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,power_law_1.2,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,power_law_1.2,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,power_law_1.2,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,power_law_1.2,0.10588159561157226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,power_law_1.2,0.12596479654312134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,power_law_1.2,0.17125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,power_law_1.2,0.21181440353393555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,power_law_1.2,0.26319360733032227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,power_law_1.2,0.34650239944458006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,power_law_1.01,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,power_law_1.01,0.1050495982170105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,power_law_1.01,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,power_law_1.01,0.10724480152130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,power_law_1.01,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,power_law_1.01,0.08922240138053894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,power_law_1.01,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,power_law_1.01,0.10770560503005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,power_law_1.01,0.11363199949264527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,power_law_1.01,0.10363520383834839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,power_law_1.01,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,power_law_1.01,0.10751999616622925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,power_law_1.01,0.11338880062103271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,power_law_1.2,0.5692800045013428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,power_law_1.01,0.11458560228347778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,power_law_1.01,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,power_law_1.01,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,power_law_1.01,0.1293887972831726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,power_law_1.01,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,power_law_1.01,0.14880640506744386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,power_law_1.01,0.17418880462646485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,power_law_1.01,0.2056960105895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,power_law_1.01,0.26162559986114503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,power_law_1.01,0.31721599102020265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,power_law_1.01,0.4365824222564697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,power_law_1.01,0.5594560146331787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,power_law_1.01,0.7938560009002685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,power_law_1.01,0.9723456382751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,balanced,0.04541333516438802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,balanced,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,balanced,0.0956053336461385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,balanced,0.09407466650009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,balanced,0.09538666407267253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,balanced,0.09603733817736308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,balanced,0.09622400005658467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,balanced,0.09650666515032451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,balanced,0.09823999802271526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,balanced,0.09939733147621155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,balanced,0.1002506713072459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,balanced,0.10043199857076009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,balanced,0.10408000151316325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,balanced,0.11174933115641277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,balanced,0.11370666821797688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,balanced,0.118367999792099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,balanced,0.13641599814097086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.14618133505185446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.17441600561141968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.20962133010228476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.27900266647338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.3271733323733012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.4503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.5670346816380819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,balanced,0.8105119864145914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.0575306415557861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,power_law_1.2,0.6348671913146973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,balanced,0.04611733555793762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,balanced,0.05459733307361603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,balanced,0.06021333237489065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,balanced,0.06587199866771698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,balanced,0.0803413341442744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,balanced,0.09299199779828389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,balanced,0.11868266264597575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,balanced,0.1423413356145223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,balanced,0.18970133860905966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,balanced,0.2280906637509664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,balanced,0.3134880065917969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,balanced,0.4030880133310954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.01,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.01,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.01,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.01,0.07930880188941955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.01,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.01,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.01,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.01,0.08899840116500854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.01,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.01,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.01,0.10320639610290527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.01,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.01,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.01,0.165555202960968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.01,0.1992192029953003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.01,0.2591295957565308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.01,0.2686847925186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,power_law_1.01,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.01,0.34828801155090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.01,0.45316481590270996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.01,0.6280511856079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,power_law_1.01,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.01,0.7932032108306885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,power_law_1.01,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,power_law_1.01,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,power_law_1.01,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,power_law_1.01,0.10881919860839843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,power_law_1.01,0.12339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,power_law_1.01,0.15940480232238768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,power_law_1.01,0.1776255965232849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,power_law_1.01,0.24599039554595947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,power_law_1.01,0.3257024049758911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,power_law_1.2,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,power_law_1.2,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,power_law_1.2,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,power_law_1.2,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,power_law_1.2,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,power_law_1.2,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,power_law_1.2,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,power_law_1.2,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,power_law_1.2,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,power_law_1.2,0.1685312032699585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,power_law_1.2,0.2482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,power_law_1.2,0.277728009223938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,power_law_1.2,0.4417856216430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,power_law_1.2,0.6086592197418212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,power_law_1.01,0.4327871799468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,power_law_1.01,0.5190976142883301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,power_law_1.01,0.79617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,power_law_1.01,1.084761619567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.08110719919204712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.12430720329284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.12336000204086303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.13392640352249147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.13355519771575927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.13868160247802735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.13882880210876464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.14269440174102782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.17123199701309205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.19249279499053956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.2368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.2881088018417358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.3896192073822021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,power_law_1.2,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.4923711776733398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,power_law_1.2,0.14481279850006104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,power_law_1.2,0.18477439880371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,power_law_1.2,0.2739264011383057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,power_law_1.2,0.3534656047821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,power_law_1.2,0.4371776103973389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,power_law_1.2,0.6282559871673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,power_law_1.2,0.6431359767913818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,power_law_1.2,0.7067776203155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,power_law_1.2,0.6692863941192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,power_law_1.2,0.6920896053314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,power_law_1.2,0.7604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,power_law_1.2,0.7440767765045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,power_law_1.2,0.7615104198455811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,power_law_1.2,0.7861695766448975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,power_law_1.2,0.7978240013122558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,power_law_1.2,0.8535360336303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,power_law_1.2,0.9651455879211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,power_law_1.2,0.9415936470031738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,power_law_1.2,1.0704768180847168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,power_law_1.2,1.1189824104309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,power_law_1.2,1.303059196472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,power_law_1.2,1.5076607704162597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,power_law_1.2,1.9924352645874024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.6985536098480225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,power_law_1.2,2.240127944946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,power_law_1.2,2.752332878112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,power_law_1.2,3.584230422973633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,balanced,0.052517334620157875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,balanced,0.08915733297665913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,balanced,0.13160533706347147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,balanced,0.2007360061009725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,balanced,0.26265066862106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,balanced,0.2670133312543233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,balanced,0.26666667064030963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,balanced,0.26821333169937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,balanced,0.269322673479716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,balanced,0.2727253238360087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,balanced,0.2762453357378642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,balanced,0.28016533454259235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,balanced,0.2837386727333069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,balanced,0.2935733397801717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,balanced,0.29711467027664185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,balanced,0.3116106589635213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,balanced,0.3582506577173869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,balanced,0.359605352083842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,balanced,0.48955734570821124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,balanced,0.5030453205108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,balanced,0.7485546271006266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,balanced,0.7725173632303873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,balanced,1.1406293710072835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,balanced,1.290175994237264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,balanced,1.9070240656534831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,balanced,2.355397383371989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.8966272354125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,1.299295997619629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.10265599489212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.1442304015159607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.1782271981239319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.19414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.19135359525680543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.19421440362930298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.19304959774017333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.18936959505081177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.19980159997940064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.21152000427246093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.21064960956573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.23070080280303956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,1.682156753540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.23416318893432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.22838399410247803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.2633280038833618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.2764287948608398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.32057600021362304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.3277695894241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,power_law_1.2,0.4286975860595703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,power_law_1.2,0.4811391830444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,power_law_1.2,0.6146431922912597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,power_law_1.2,0.7571712017059327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,power_law_1.2,1.1115072250366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,power_law_1.2,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,power_law_1.2,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,power_law_1.2,0.18942079544067383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.0765504002571106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,power_law_1.2,0.32314879894256593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,power_law_1.2,0.3321536064147949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,power_law_1.2,0.3465471982955933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,power_law_1.2,0.3540735960006714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,power_law_1.2,0.3660991907119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,power_law_1.2,0.3675839900970459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,power_law_1.2,0.38210558891296387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,power_law_1.2,0.389683198928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,power_law_1.2,0.3995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,power_law_1.2,0.42368640899658205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,power_law_1.2,0.4397183895111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,power_law_1.2,0.4369408130645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,power_law_1.2,0.49472641944885254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,power_law_1.2,0.5240831851959229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,power_law_1.2,0.6148992061614991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,power_law_1.2,0.6706175804138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,power_law_1.2,0.8250176429748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,power_law_1.2,0.9878463745117188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.12059520483016968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,power_law_1.2,1.321548843383789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,power_law_1.2,1.6632448196411134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,power_law_1.2,1.383135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.19575040340423583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,power_law_1.2,2.2987648010253907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.12492159605026246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.10928000211715698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.1115839958190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.1085055947303772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.11706240177154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.1225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.1304255962371826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.16031359434127807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.1897472023963928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.21081600189208985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.2456831932067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.32575359344482424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.4194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.5486656188964844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.7179327964782715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,1.0239744186401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,power_law_1.2,2.9246847152709963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.35526399612426757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.41992959976196287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.2,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.2,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.2,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.2,0.053299200534820554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.2,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.2,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.2,0.0626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.2,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.2,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.2,0.13404159545898436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.2,0.14702080488204955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.2,0.2043519973754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.2,0.2221247911453247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.2,0.3113215923309326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.2,0.34990720748901366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.2,0.49816322326660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.2,0.6116352081298828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.07870720028877258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.14328320026397706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.18561919927597045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.21635839939117432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.28958079814910886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.2958208084106445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.30408320426940916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.318668794631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.3412031888961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.3440448045730591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.3612864017486572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.3773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.3831615924835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.40654721260070803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.46460800170898436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.5171775817871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.6052864074707032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.6923327922821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.8562239646911621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.068832015991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.3818367958068847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.792755126953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.416332817077637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,power_law_1.2,2.950182342529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,power_law_1.2,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,power_law_1.2,0.06961280107498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,power_law_1.2,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,power_law_1.2,0.09379839897155762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,power_law_1.2,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,power_law_1.2,0.10100480318069457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,power_law_1.2,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,power_law_1.2,0.11523840427398682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,power_law_1.2,0.12088960409164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,power_law_1.2,0.12488319873809814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,power_law_1.2,0.13422080278396606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,power_law_1.2,0.14010239839553834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,power_law_1.2,0.15158400535583497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,power_law_1.2,0.18435200452804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,power_law_1.2,0.21025919914245605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,power_law_1.2,0.26657919883728026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,power_law_1.2,0.32351999282836913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,power_law_1.2,0.436294412612915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,power_law_1.2,0.5517248153686524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,power_law_1.2,0.7786111831665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,power_law_1.2,0.9948863983154297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,power_law_1.2,1.4315967559814453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,power_law_1.2,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,power_law_1.2,0.06035839915275574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,power_law_1.2,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,power_law_1.2,0.062636798620224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,power_law_1.2,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,power_law_1.2,0.08260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,power_law_1.2,0.08678399920463561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,power_law_1.2,0.12370560169219971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.2,0.1434623956680298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.2,0.18810240030288697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,power_law_1.2,1.841689682006836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.2,0.23283839225769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.2,0.3128576040267944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.2,0.4101503849029541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,balanced,0.0730453332265218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,balanced,0.11563733220100403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,balanced,0.11853333314259847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,balanced,0.11930132905642192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,balanced,0.11978666981061299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,balanced,0.12219199538230896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,balanced,0.12300266822179158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,balanced,0.12007466952006023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,balanced,0.12104533116022746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,balanced,0.12507733702659607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,balanced,0.1239466667175293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,balanced,0.1255626678466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,balanced,0.1313706636428833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,balanced,0.1420746644337972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,balanced,0.14406399925549826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,balanced,0.14562132954597473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,balanced,0.17495467265446982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,balanced,0.1737013260523478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,balanced,0.23455466826756796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,balanced,0.26506133874257404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,balanced,0.3465119997660319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,balanced,0.40296534697214764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,balanced,0.5742293198903402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,balanced,0.7370186646779379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,balanced,1.0607840220133464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,balanced,1.3766026496887207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.2,0.6393343925476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.2,0.7577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.2,1.1535936355590821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.07160959839820862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.2,1.46113920211792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.10837119817733765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.13464959859848022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.17760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,balanced,0.09267200032869975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,balanced,0.04756266872088114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,balanced,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,balanced,0.09865066409111023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,balanced,0.09757866462071736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,balanced,0.09857599933942159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,balanced,0.0995786686738332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,balanced,0.1002400020758311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.24024319648742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,balanced,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,balanced,0.10204799969991048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,balanced,0.10263466835021973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,balanced,0.10804266730944316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,balanced,0.10712533195813496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,balanced,0.10956799983978271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,balanced,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,balanced,0.1181653340657552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,balanced,0.12677866220474243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,balanced,0.13582932949066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,balanced,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,balanced,0.17460266749064127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,balanced,0.2042293349901835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,balanced,0.2420426607131958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,balanced,0.31989334026972455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,balanced,0.39112532138824463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.3180032014846802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,power_law_1.01,0.38394238948822024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,balanced,0.039349332451820374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,balanced,0.0393653338154157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,balanced,0.043338666359583534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,balanced,0.05295999844868978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,balanced,0.07533333202203114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,balanced,0.09690133730570476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,balanced,0.10059199730555217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,balanced,0.12821867068608603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,balanced,0.14352533221244812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,balanced,0.19406400124231973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,balanced,0.23626132806142172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,balanced,0.330458660920461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,balanced,0.4241653283437093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,power_law_1.01,0.5348608016967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,power_law_1.01,0.6891776084899902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,power_law_1.01,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,power_law_1.01,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,power_law_1.01,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,power_law_1.01,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,power_law_1.01,0.06949759721755981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,power_law_1.01,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,power_law_1.01,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,power_law_1.01,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,power_law_1.01,0.09147520065307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.01,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.01,0.11563520431518555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.01,0.1407871961593628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.01,0.16917120218276976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.01,0.2088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.01,0.2883584022521973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.01,0.33987839221954347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.01,0.5128320217132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.01,0.6994112014770508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,power_law_1.2,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,power_law_1.2,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,power_law_1.2,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,power_law_1.2,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,power_law_1.2,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,power_law_1.2,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,power_law_1.2,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,power_law_1.2,0.06536960005760192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,power_law_1.2,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,power_law_1.2,0.11982719898223877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,power_law_1.2,0.14479999542236327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,power_law_1.2,0.18963199853897095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,power_law_1.2,0.24747519493103026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,power_law_1.2,0.35395839214324953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,power_law_1.2,0.4305856227874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,power_law_1.01,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,power_law_1.01,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,power_law_1.01,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,power_law_1.01,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,power_law_1.01,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,power_law_1.01,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,power_law_1.01,0.07402240037918091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,power_law_1.01,0.08222079873085023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,power_law_1.01,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.01,0.12067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.01,0.14369920492172242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.01,0.18672000169754027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.01,0.2358464002609253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.01,0.3180608034133911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.01,0.41884799003601075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.01,0.5888192176818847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.01,0.8274304389953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.09827839732170104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,power_law_1.2,0.6621119976043701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.13719680309295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.17386879920959472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.2133568048477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,power_law_1.2,0.8980863571166993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.2961280107498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.3758975982666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,power_law_1.2,0.06600319743156433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,power_law_1.2,0.08218240141868591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,power_law_1.2,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,power_law_1.2,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,power_law_1.2,0.10056960582733154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,power_law_1.2,0.09434239864349366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,power_law_1.2,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,power_law_1.2,0.10202239751815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,power_law_1.2,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,power_law_1.2,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,power_law_1.2,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,power_law_1.2,0.10688639879226684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,power_law_1.2,0.11128959655761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.5390336036682128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,power_law_1.2,0.11759359836578369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,power_law_1.2,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,power_law_1.2,0.12432639598846436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,power_law_1.2,0.13982720375061036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.2,0.15444480180740355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.2,0.18294399976730347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.2,0.20392959117889403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.2,0.3185215950012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.2,0.38103680610656737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.2,0.5277247905731202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.2,0.6356800079345704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.2,1.1086976051330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.2,1.3981504440307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,balanced,0.029733332494894665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,balanced,0.029765332738558452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,balanced,0.05099200208981832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,balanced,0.0781173308690389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,balanced,0.07903466622034709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,balanced,0.07982400059700012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,balanced,0.07921599845091502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,balanced,0.07948799928029378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,balanced,0.0801333338022232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,balanced,0.08796800176302592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,balanced,0.09074133634567261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,balanced,0.09340266386667888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,balanced,0.09768533706665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,balanced,0.10493333141009013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,balanced,0.11079999804496765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,balanced,0.12801067034403482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,balanced,0.13760000467300415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,balanced,0.17339734236399332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,balanced,0.19572800397872925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,balanced,0.27618134021759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,balanced,0.34552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,balanced,0.4788159926732381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,balanced,0.6173493464787801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.6937215805053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,balanced,0.042394667863845825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,balanced,0.05809600154558817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,balanced,0.06244266529877981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,balanced,0.062362665931383766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,balanced,0.06621333460013072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,balanced,0.0703893353541692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,balanced,0.07187200089295705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,balanced,0.0784693310658137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,balanced,0.09125866492589314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,balanced,0.104912002881368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,balanced,0.13076266646385193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,balanced,0.1506666640440623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,balanced,0.19050133228302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,balanced,0.2421813408533732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,balanced,0.32337067524592084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,balanced,0.40273598829905194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,balanced,0.5797813336054484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,balanced,0.7430720329284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,power_law_1.01,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,power_law_1.01,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,power_law_1.01,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,power_law_1.01,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,power_law_1.01,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,power_law_1.01,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,power_law_1.01,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,power_law_1.01,0.08679680228233337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,power_law_1.01,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,power_law_1.01,0.11741440296173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,power_law_1.01,0.12074240446090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,power_law_1.01,0.13997440338134765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,power_law_1.01,0.152729594707489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,power_law_1.01,0.1938431978225708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,power_law_1.01,0.2324608087539673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,power_law_1.01,0.30753920078277586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,power_law_1.01,0.3665087938308716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,power_law_1.01,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,power_law_1.01,0.5345536231994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,power_law_1.01,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,power_law_1.01,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,power_law_1.01,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,power_law_1.01,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,power_law_1.01,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,power_law_1.01,0.06799359917640686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,power_law_1.01,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,power_law_1.01,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,power_law_1.01,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,power_law_1.01,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,power_law_1.01,0.13408639430999755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,power_law_1.01,0.1506175994873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,power_law_1.01,0.19730559587478638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,power_law_1.01,0.25834240913391116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,power_law_1.01,0.3469504117965698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,power_law_1.01,0.45907201766967776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,power_law_1.01,0.6969535827636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,power_law_1.01,0.890835189819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,power_law_1.01,0.6428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,power_law_1.01,0.9549504280090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.10152319669723511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.11450239419937133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.12307840585708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.12220799922943115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.1276095986366272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.13006080389022828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.13525760173797607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.1387392044067383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.14080640077590942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.1473536014556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.15360000133514404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.18202240467071534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.20798079967498778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.24805119037628173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.25647358894348143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.343507194519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.4267583847045898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.5311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.6515135765075684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,0.9420479774475098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,power_law_1.2,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,power_law_1.2,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,power_law_1.2,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,power_law_1.2,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,power_law_1.2,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,power_law_1.2,0.18055039644241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,power_law_1.2,0.22072958946228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,power_law_1.2,0.32124159336090086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,power_law_1.2,0.3816832065582275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,power_law_1.2,0.6115903854370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,power_law_1.01,1.2708736419677735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,power_law_1.2,0.7090879917144776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.2454400062561035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.10696320533752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.1654080033302307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.17623679637908934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.19912320375442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.20478720664978028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.21413760185241698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.2166208028793335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.22834560871124268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.22607359886169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.23598721027374267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.24312961101531982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.2562560081481934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.25647358894348143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.29244160652160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.2978111982345581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.3648000001907349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.3919680118560791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.45934081077575684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,power_law_1.01,0.5175487995147705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,power_law_1.01,0.6532671928405762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,power_law_1.01,0.8074560165405273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,power_law_1.01,1.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,power_law_1.01,1.3323712348937988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.2,0.180511999130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.2,0.2425856113433838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.2,0.2510143995285034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.2,0.34609920978546144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.2,0.3913856029510498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.2,0.3800640106201172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.2,0.39028480052948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.2,0.3842175960540771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.2,0.3963776111602783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.2,0.3953344106674194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.2,0.4018496036529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.2,0.43091840744018556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.2,0.4432191848754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.2,0.4434175968170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.2,0.45810561180114745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.2,0.477023983001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.2,0.504095983505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.2,0.5215360164642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.2,0.5957824230194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.2,0.7093120098114014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.2,0.795033597946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.2,1.006118392944336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.2,0.9398271560668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.2,1.1741888046264648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.2,1.523750400543213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.029382398724555968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.02951039969921112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.11286400556564331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.10882560014724732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.12892800569534302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.13271679878234863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.14394880533218385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.14048000574111938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.13408639430999755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.14277119636535646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.15492479801177977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.2,2.0488000869750977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.15633280277252198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.17125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.1850559949874878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.2003391981124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.22644479274749757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.2624768018722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.333407998085022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.38229761123657224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.48803200721740725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.6039103984832763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.21094400882720948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,0.8418751716613769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,1.0790207862854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.24959359169006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.2,2.489683151245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.3646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.5445055961608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,power_law_1.2,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,power_law_1.2,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,power_law_1.2,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,power_law_1.2,0.07615360021591186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,power_law_1.2,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,power_law_1.2,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,power_law_1.2,0.07804160118103028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,power_law_1.2,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,power_law_1.2,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,power_law_1.2,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,power_law_1.2,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,power_law_1.2,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.2,0.1292415976524353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.2,0.15997439622879028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.2,0.18984960317611693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.2,0.2944511890411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.2,0.35868799686431885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.2,0.5216063976287841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.2,0.7224063873291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.2,1.0479935646057128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.2,1.3159168243408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.0779583990573883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.14396159648895263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,power_law_1.01,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.14860160350799562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.14535679817199706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.15852160453796388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,power_law_1.01,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.16035200357437135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.16962560415267944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.18332159519195557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.18775039911270142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.2009023904800415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.2078336000442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.22327680587768556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.2605056047439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.26880640983581544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,power_law_1.01,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.32316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.35479040145874025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.45041918754577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.5432064056396484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,power_law_1.2,0.7132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,power_law_1.2,0.9265279769897461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,power_law_1.2,1.2982080459594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,power_law_1.2,1.6895423889160157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.01,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.01,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.01,0.12827520370483397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.01,0.19882880449295043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.01,0.2681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,power_law_1.01,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,power_law_1.01,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,power_law_1.01,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,power_law_1.01,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,power_law_1.01,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,power_law_1.01,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.01,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.01,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.01,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.01,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.01,0.13582719564437867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.01,0.17964799404144288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.01,0.22592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.01,0.31715199947357176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.01,0.41901440620422364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.01,0.38426880836486815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.01,0.49681921005249025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.01,0.833568000793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.01,1.052889633178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,balanced,0.05029866596062978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,balanced,0.051669334371884666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,balanced,0.06851733227570851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,balanced,0.08708799878756206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,balanced,0.09013866384824117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,balanced,0.0922933320204417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,balanced,0.08990933497746785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,balanced,0.09195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,balanced,0.09170666337013245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,balanced,0.09275199969609578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,balanced,0.09436266620953877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,balanced,0.09499200185139973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,balanced,0.10499200224876404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,balanced,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,balanced,0.10968533158302307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,balanced,0.11957333485285442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,balanced,0.12406399846076965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,balanced,0.14145066340764365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,balanced,0.15430933237075806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,balanced,0.18812799453735352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,balanced,0.21420266230901083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,balanced,0.29811733961105347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,balanced,0.36314133803049725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,balanced,0.5042879978815714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,balanced,0.6634026765823364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.04476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.1272704005241394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.16488319635391235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.2189568042755127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,power_law_1.01,0.3099263906478882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,power_law_1.01,0.4088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,power_law_1.01,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,power_law_1.01,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,power_law_1.01,0.0584384024143219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,power_law_1.01,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,power_law_1.01,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,power_law_1.01,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,power_law_1.01,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,power_law_1.01,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,power_law_1.01,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,power_law_1.01,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,power_law_1.01,0.1078336000442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,power_law_1.01,0.12172160148620606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,power_law_1.01,0.13101439476013182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,power_law_1.01,0.15580159425735474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,power_law_1.01,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,power_law_1.01,0.226476788520813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,power_law_1.01,0.2723072052001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,power_law_1.01,0.3698816061019897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,power_law_1.01,0.5157695770263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,power_law_1.01,0.682144021987915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,power_law_1.01,0.9057536125183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,balanced,0.04410133262475332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,balanced,0.04461333155632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,balanced,0.07237333556016286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,balanced,0.10146666566530864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,balanced,0.13146666685740152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,balanced,0.1290079951286316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,balanced,0.13035733501116434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,balanced,0.1301813324292501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,balanced,0.13301333785057068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,balanced,0.13479466239611307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,balanced,0.13618133465449014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,balanced,0.13577600320180258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,balanced,0.14221866925557455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,balanced,0.14275200168291727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,balanced,0.15153599778811136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,balanced,0.15846932927767435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,balanced,0.1687893271446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,balanced,0.18725866079330444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,balanced,0.20879999796549478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,balanced,0.24503467480341592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,balanced,0.2902453343073527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,balanced,0.3622719844182332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,balanced,0.44627201557159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,balanced,0.5981119871139526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,balanced,0.7463839848836263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,balanced,0.05390933156013489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,balanced,0.06886933247248332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,balanced,0.09892800450325012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,balanced,0.13685333728790283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,balanced,0.1402613321940104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,balanced,0.14128533005714417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,balanced,0.142085333665212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,balanced,0.14269333084424338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,balanced,0.14542399843533835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,balanced,0.1469066639741262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,balanced,0.14730133612950644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,balanced,0.15015467007954916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,balanced,0.15761066476504007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,balanced,0.160453329483668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,balanced,0.16446933150291443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,balanced,0.1754080057144165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.18767466147740683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.21478400627772012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.23734400669733682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.28988800446192425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.33537065982818604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.4729546705881755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.572869340578715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.8208373387654623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,balanced,1.0506880283355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,power_law_1.01,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,power_law_1.01,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,power_law_1.01,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,power_law_1.01,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,power_law_1.01,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,power_law_1.01,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,power_law_1.01,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,power_law_1.01,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,power_law_1.01,0.05809280276298523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,power_law_1.01,0.09211519956588746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,power_law_1.01,0.11630719900131226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,power_law_1.01,0.142848002910614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,power_law_1.01,0.2011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,power_law_1.01,0.25141119956970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,power_law_1.01,0.3464832067489624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,power_law_1.01,0.45908479690551757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,4,power_law_1.01,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,4,power_law_1.01,0.1552575945854187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,4,power_law_1.01,0.2090303897857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,4,power_law_1.01,0.20419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,4,power_law_1.01,0.2535423994064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,4,power_law_1.01,0.2921087980270386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,4,power_law_1.01,0.3236736059188843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,4,power_law_1.01,0.3845184087753296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,4,power_law_1.01,0.36753919124603274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,4,power_law_1.01,0.3889152050018311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,4,power_law_1.01,0.3484992027282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,4,power_law_1.01,0.3864255905151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,4,power_law_1.01,0.4097919940948486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,4,power_law_1.01,0.40144639015197753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,4,power_law_1.01,0.3957760095596313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,4,power_law_1.01,0.40592641830444337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,4,power_law_1.01,0.42938880920410155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,4,power_law_1.01,0.4342336177825928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.01,0.46378240585327146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.028518399596214293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.01,0.5197440147399902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.01,0.5549952030181885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.01,0.6509312152862549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.01,0.79617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.01,1.01527681350708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.01,1.262553596496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.01,1.8163263320922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.01,2.338700866699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.1241536021232605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.1407807946205139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.1926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.21772799491882325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.40074877738952636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,power_law_1.01,0.43639678955078126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,balanced,0.05740800003210703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,balanced,0.06398400167624156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,balanced,0.06459199885527293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,balanced,0.06490666667620341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,balanced,0.06479999919732411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,balanced,0.06713599960009257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,balanced,0.06857066849867503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,balanced,0.07870933413505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,balanced,0.08415466547012329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,balanced,0.10777067144711812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,balanced,0.10010666648546855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,balanced,0.1311893363793691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,balanced,0.14658666650454202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,balanced,0.19963733355204263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,balanced,0.25442665815353394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,balanced,0.36956266562143963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,balanced,0.48308265209198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,balanced,0.6913066705067953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,balanced,0.8975786368052164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,balanced,0.05680533250172933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,balanced,0.07982933521270752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,balanced,0.13531733552614847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,balanced,0.13433600465456644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,balanced,0.1360266705354055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,balanced,0.13758933544158936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,balanced,0.13962666193644205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,balanced,0.13749866684277853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,balanced,0.13819733262062073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,balanced,0.13734933733940125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,balanced,0.14106667041778564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,balanced,0.13979199528694153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,balanced,0.1418719987074534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,balanced,0.14800000190734863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,balanced,0.15147733688354492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,balanced,0.15434666474660239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,balanced,0.1657546659310659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,balanced,0.18969066937764487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,balanced,0.19884800910949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,balanced,0.24240533510843912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,balanced,0.2879466613133748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,balanced,0.37594131628672284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,balanced,0.4410933256149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,balanced,0.6261866490046183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,balanced,0.7867733637491862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,balanced,1.1658506393432617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,balanced,1.520085334777832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.07136639952659607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.07898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.115065598487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.14318079948425294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.1641983985900879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.2000960111618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.24774401187896727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.34167680740356443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.4335296154022217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.5879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.7520127773284913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.07526400089263915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.08863360285758973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.08995839953422546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.11502079963684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.14573440551757813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.1770624041557312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.2362816095352173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.2832767963409424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.3992448091506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.5002816200256348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,power_law_1.01,0.6624703884124756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,power_law_1.01,0.8696127891540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,balanced,0.042954668402671814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,balanced,0.04494933287302653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,balanced,0.051301335295041404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,balanced,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,balanced,0.09714133540789287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,balanced,0.09667733311653137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,balanced,0.09662399689356486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,balanced,0.09798933068911235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,balanced,0.09689066807428996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,balanced,0.09737066427866618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,balanced,0.10075733065605164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,balanced,0.10110933581988017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,balanced,0.10178666313489278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,balanced,0.10925333698590596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,balanced,0.10925867160161336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,balanced,0.11430933078130086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,balanced,0.12729600071907043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,balanced,0.13915733496348062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,balanced,0.1590506633122762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,balanced,0.17726399501164755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,balanced,0.2534079949061076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,balanced,0.2876053253809611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,balanced,0.4172159830729167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,balanced,0.5115893284479777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,balanced,0.7538932959238688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,balanced,0.9783039887746176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,balanced,0.027679999669392902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,balanced,0.02940266579389572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,balanced,0.03396799912055334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,balanced,0.06312533219655354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,balanced,0.07461333274841309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,balanced,0.08408000071843465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,balanced,0.10101866722106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,balanced,0.1260586678981781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,balanced,0.15424000223477682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,balanced,0.19196800390879312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,8,power_law_1.01,0.12874879837036132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,8,power_law_1.01,0.18702080249786376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,8,power_law_1.01,0.2969280004501343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,8,power_law_1.01,0.2805567979812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,8,power_law_1.01,0.33678719997406004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,8,power_law_1.01,0.3637248039245605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,8,power_law_1.01,0.33000319004058837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,8,power_law_1.01,0.3933824062347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,8,power_law_1.01,0.3804480075836182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,8,power_law_1.01,0.39163520336151125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,8,power_law_1.01,0.36579840183258056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,8,power_law_1.01,0.3811392068862915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,8,power_law_1.01,0.3902143955230713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,8,power_law_1.01,0.35923840999603274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,8,power_law_1.01,0.3914175987243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,8,power_law_1.01,0.36241281032562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,8,power_law_1.01,0.3826303958892822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,8,power_law_1.01,0.4210239887237549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.01,0.4327871799468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.01,0.47176318168640136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.01,0.513753604888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.01,0.6092095851898194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.01,0.699238395690918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.01,0.912019157409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.01,1.1215423583984374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.01,1.5773311614990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.01,1.9558847427368165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.08301439881324768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.09626240134239197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.10724480152130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.1053056001663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.10927360057830811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.11329280138015747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.13115520477294923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.16457600593566896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.20419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.23495039939880372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.30192639827728274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.3975744009017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.52674560546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.664569616317749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,power_law_1.2,0.9424448013305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.1684864044189454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,balanced,0.0689333329598109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,balanced,0.08270399769147237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,balanced,0.08331199983755748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,balanced,0.08125333487987518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,balanced,0.08345599969228108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,balanced,0.08667733271916707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,balanced,0.08721599976221721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,balanced,0.0853653351465861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,balanced,0.08946133653322856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,balanced,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,balanced,0.09616000453631084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,balanced,0.10339732964833577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,balanced,0.11184533437093098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,balanced,0.12217066685358684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,balanced,0.14587199687957764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,balanced,0.16433067123095194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,balanced,0.20520534118016562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,balanced,0.2454986572265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,balanced,0.3274773359298706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,balanced,0.4219893217086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,balanced,0.5865813493728638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,balanced,0.7511146863301595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,8,power_law_1.2,0.1291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,8,power_law_1.2,0.18706560134887695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,8,power_law_1.2,0.24757120609283448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,8,power_law_1.2,0.26369919776916506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,8,power_law_1.2,0.30033919811248777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,8,power_law_1.2,0.3761791944503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,8,power_law_1.2,0.36285440921783446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,8,power_law_1.2,0.341644811630249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,8,power_law_1.2,0.3922816038131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,8,power_law_1.2,0.3351423978805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,8,power_law_1.2,0.37079041004180907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,8,power_law_1.2,0.3612351894378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,8,power_law_1.2,0.3603264093399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,8,power_law_1.2,0.3625663995742798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,8,power_law_1.2,0.38622078895568845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,8,power_law_1.2,0.3921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,8,power_law_1.2,0.39566080570220946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,8,power_law_1.2,0.42696318626403806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.2,0.43493118286132815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.2,0.48714241981506345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.2,0.5464960098266601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.2,0.6818111896514892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.2,0.8047103881835938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.2,0.983244800567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.2,1.2632320404052735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.2,1.8166143417358398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.2,2.1237823486328127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,balanced,0.039690665900707245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,balanced,0.04218666752179464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,balanced,0.04240000247955322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,balanced,0.04398933549722036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,balanced,0.046869332591692604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,balanced,0.04808000226815542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,balanced,0.0787306676308314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,balanced,0.0897920032342275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,balanced,0.11167466640472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,balanced,0.1320746640364329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.07029759883880615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.09906560182571411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.10740480422973633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.12055039405822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.1504256010055542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.17191679477691652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.20559999942779542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.27520639896392823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.3558207988739014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.4763455867767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.22243199348449708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.6342016220092773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,power_law_1.01,0.30899839401245116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,power_law_1.01,0.3598848104476929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,0.9118016242980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,power_law_1.01,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,power_law_1.01,0.11195520162582398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,power_law_1.01,0.11612160205841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,power_law_1.01,0.11951359510421752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,power_law_1.01,0.12195839881896972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,power_law_1.01,0.12490240335464478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,power_law_1.01,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,power_law_1.01,0.13377280235290528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,power_law_1.01,0.13553279638290405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,power_law_1.01,0.1440832018852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,power_law_1.01,0.1553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,power_law_1.01,0.16091519594192505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,power_law_1.01,0.17059839963912965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,power_law_1.01,0.1951807975769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,power_law_1.01,0.21483519077301025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,power_law_1.01,0.26590719223022463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,power_law_1.01,0.3053760051727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,power_law_1.01,0.40879359245300295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,power_law_1.01,0.5044288158416748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,power_law_1.01,0.6945151805877685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,power_law_1.01,0.8861184120178223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,power_law_1.01,1.2641728401184082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,power_law_1.01,1.672172737121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,power_law_1.01,0.5336448192596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,balanced,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,balanced,0.0802346666653951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,balanced,0.08110400040944417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,balanced,0.07955199976762135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,balanced,0.0822026679913203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,balanced,0.08285333216190338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,balanced,0.08358933528264363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,balanced,0.0846560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,balanced,0.08499200145403545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,balanced,0.09060800075531006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,balanced,0.09489599863688152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,balanced,0.10626133282979329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,balanced,0.10955733060836792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,balanced,0.13499200344085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,balanced,0.1516266663869222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,balanced,0.19722666343053183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,balanced,0.1812373399734497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,balanced,0.23799467086791992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,balanced,0.26415467262268066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,balanced,0.3607306480407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,balanced,0.4144959847132365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,power_law_1.01,0.7546048164367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.046668815612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,balanced,0.04678933322429657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,balanced,0.050245334704717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,balanced,0.05562133093674978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,balanced,0.054586668809254967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,balanced,0.06864533325036366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,balanced,0.07067200044790904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,balanced,0.07452799876530965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,balanced,0.08458667000134786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,balanced,0.09584533174832661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,balanced,0.10698666175206502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,balanced,0.13361600041389465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,balanced,0.15436800320943198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,balanced,0.20138667027155557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,balanced,0.24555200338363647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,balanced,0.3465813398361206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,balanced,0.44125866889953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,balanced,0.03605866680542628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,balanced,0.037946666280428566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,balanced,0.04541333516438802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,balanced,0.061674664417902626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,balanced,0.06814933319886525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,balanced,0.07855999966462453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,balanced,0.09850133458773296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,balanced,0.10925867160161336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,balanced,0.146506667137146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,balanced,0.18330667416254678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,balanced,0.2504853407541911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,balanced,0.3264533281326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,balanced,0.47009599208831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,balanced,0.6200160185496012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.0748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.09377920031547546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.09711359739303589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.10033919811248779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.11060479879379273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.12049920558929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.1539903998374939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.1923135995864868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.20266880989074706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.259500789642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.3157504081726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.4307136058807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.564569616317749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.7281599998474121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,0.9385151863098145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,power_law_1.2,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,power_law_1.2,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,power_law_1.2,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,power_law_1.2,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,power_law_1.2,0.111571204662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,power_law_1.2,0.14394880533218385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,power_law_1.2,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,power_law_1.2,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,power_law_1.2,0.15461119413375854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,power_law_1.2,0.15409280061721803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,power_law_1.2,0.1571903944015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,power_law_1.2,0.16640000343322753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,power_law_1.2,0.17142399549484252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,power_law_1.2,0.18032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,power_law_1.2,0.1812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,power_law_1.2,0.19440000057220458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,power_law_1.2,0.21775360107421876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,power_law_1.2,0.23338239192962645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,power_law_1.2,0.2759936094284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,power_law_1.2,0.29599359035491946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,power_law_1.2,0.37141759395599366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,power_law_1.2,0.42230401039123533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,power_law_1.2,0.5376704216003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,power_law_1.2,0.6669248104095459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,power_law_1.2,0.9753279685974121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,power_law_1.2,1.1432191848754882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.0974016010761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.13164160251617432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.12847360372543334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.1582335948944092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.15530879497528077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.16208640336990357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.16501760482788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.16374399662017822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.1729856014251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.182368004322052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.18970880508422852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.1907520055770874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.21492478847503663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.23066880702972412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.26743040084838865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.3056191921234131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.37868800163269045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.4517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.6047679901123046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.7512191772460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,1.0126848220825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.3910464286804198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,balanced,0.039919999738534294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,balanced,0.045082668463389076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,balanced,0.05223466455936432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,balanced,0.05369600156943003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,balanced,0.08796266714731853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,balanced,0.09699733058611552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,balanced,0.12651733557383218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,balanced,0.15245333313941956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,balanced,0.20116267601648966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,balanced,0.24688533941904703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,balanced,0.3577226797739665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,balanced,0.45843199888865155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,balanced,0.06135466694831848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,balanced,0.059338668982187905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,balanced,0.05946133534113566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,balanced,0.05958400170008341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,balanced,0.0598826656738917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,balanced,0.06154133379459381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,balanced,0.06273066500822704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,balanced,0.06366933385531108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,balanced,0.08349866668383281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,balanced,0.0800799975792567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,balanced,0.10072533289591472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,balanced,0.10785067081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,balanced,0.14062399665514627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,balanced,0.15267733732859293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,balanced,0.2131040096282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,balanced,0.24441067377726236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,balanced,0.34452799956003827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,balanced,0.44550931453704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,balanced,0.03178666780392329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,balanced,0.03419733295838038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,balanced,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,balanced,0.043375998735427856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,balanced,0.051642666260401406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,balanced,0.0747519979874293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,balanced,0.10612266262372334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,balanced,0.12585600217183432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,balanced,0.16899200280507407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,balanced,0.20522133509318033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,balanced,0.29505600531895954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,balanced,0.3685973485310872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.09168639779090881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.11550719738006592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.15771520137786865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.17939200401306152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.25464959144592286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.30513920783996584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.10969599485397338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.4850751876831055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.13587199449539183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.21249279975891114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,balanced,0.18433600664138794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,balanced,0.31838399171829224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,balanced,0.5895520051320394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,balanced,1.1246559619903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,balanced,2.1604746182759604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,balanced,3.162522633870443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,balanced,3.172783851623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,balanced,3.174000104268392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,balanced,3.1809492111206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,balanced,3.1992534001668296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,balanced,3.211306571960449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,balanced,3.2293386459350586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,balanced,3.2426932652791343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,balanced,3.267754554748535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.5404607772827148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,balanced,3.3070665995279946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,balanced,3.3273706436157227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,balanced,3.3644320170084634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,balanced,3.6805013020833335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,balanced,3.5381011962890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,balanced,4.366618792215983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,balanced,3.8874025344848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,balanced,4.87992000579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,balanced,4.5374399820963545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,balanced,7.196874618530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,balanced,6.699962615966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,balanced,11.447252909342447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,balanced,10.250410715738932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.23214080333709716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,power_law_1.2,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,power_law_1.2,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,power_law_1.2,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,power_law_1.2,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,power_law_1.2,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,power_law_1.2,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,power_law_1.2,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,power_law_1.2,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,power_law_1.2,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,power_law_1.2,0.0918720006942749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,power_law_1.2,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,power_law_1.2,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,power_law_1.2,0.13460479974746703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,power_law_1.2,0.13972480297088624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,power_law_1.2,0.16909439563751222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,power_law_1.2,0.20510079860687255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,power_law_1.2,0.2805504083633423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,power_law_1.2,0.3325376033782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,power_law_1.2,0.4880832195281982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,power_law_1.2,0.6174975872039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,power_law_1.2,0.02913280129432678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,power_law_1.2,0.02810879945755005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,power_law_1.2,0.026598399877548216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,power_law_1.2,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,power_law_1.2,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,power_law_1.2,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,power_law_1.2,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,power_law_1.2,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,power_law_1.2,0.12318719625473022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.06728960275650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,power_law_1.2,0.15998719930648803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.09855999946594238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.13049600124359131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.16303999423980714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.18533120155334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.23764479160308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.2614464044570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.38499200344085693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,power_law_1.2,0.4745215892791748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,power_law_1.2,0.7225279808044434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,power_law_1.2,0.800607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,power_law_1.2,0.262175989151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.08799359798431397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.08996480107307434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.09153280258178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.10550400018692016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.1084671974182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.12058240175247192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.1361791968345642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.15299839973449708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.2054271936416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.2664448022842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.3386431932449341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.4407872200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.5512256145477294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,power_law_1.01,0.7894400119781494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,power_law_1.01,0.9957504272460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,power_law_1.2,0.2850048065185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.14272639751434327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.19621119499206544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.24218239784240722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.3285952091217041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.40776958465576174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.39461119174957277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.5828991889953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.5829951763153076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.5508863925933838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.5416895866394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.5284607887268067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.49428482055664064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.5321856021881104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.5846079826354981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.5925055980682373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.5937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.6515520095825196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.6952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,power_law_1.2,0.8079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,power_law_1.2,0.7943615913391113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,power_law_1.2,0.930355167388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.0934528350830077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,power_law_1.2,1.3645631790161132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,power_law_1.2,1.4096192359924316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,power_law_1.2,1.8441791534423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,power_law_1.2,2.25034236907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.11434880495071412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.11283199787139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.11818879842758179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.1181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.1280832052230835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.11582080125808716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.13569920063018798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.1404736042022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.15180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.1755328059196472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.2435904026031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.2473088026046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.3075648069381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,power_law_1.01,0.315391993522644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,power_law_1.01,0.41570558547973635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,power_law_1.01,0.5285439968109131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,power_law_1.01,0.7670976161956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,power_law_1.01,0.8022720336914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.07852799892425537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.12387200593948364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.13574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,2,power_law_1.01,0.38575360774993894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,2,power_law_1.01,0.4335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,2,power_law_1.01,0.4875840187072754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,2,power_law_1.01,0.6774847984313965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,2,power_law_1.01,0.8627776145935059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.176147198677063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,2,power_law_1.01,1.204255962371826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,2,power_law_1.01,1.301363182067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,2,power_law_1.01,1.3309951782226563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,2,power_law_1.01,1.3436479568481445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,2,power_law_1.01,1.4462335586547852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,2,power_law_1.01,1.4042688369750977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,2,power_law_1.01,1.4461567878723145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,2,power_law_1.01,1.4561216354370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,2,power_law_1.01,1.4762240409851075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,2,power_law_1.01,1.49552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,2,power_law_1.01,1.5147263526916503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,2,power_law_1.01,1.5448512077331542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,2,power_law_1.01,1.6328575134277343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.01,1.6674432754516602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.01,1.8887359619140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.01,2.102956771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.01,2.579520034790039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.01,2.9471551895141603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.01,4.046015930175781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.01,4.9282176971435545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.01,7.008147430419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.22269439697265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.01,8.831366729736327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.31024000644683836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,balanced,0.06474133332570393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,balanced,0.10604799787203471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,balanced,0.14974400401115417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,balanced,0.22708266973495483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,balanced,0.30664000908533734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,balanced,0.30989332993825275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,balanced,0.3115946650505066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,balanced,0.31705600023269653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,balanced,0.3144853313763936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,balanced,0.3160853385925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,balanced,0.3140160044034322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,balanced,0.320357342561086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,balanced,0.3280319968859355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,balanced,0.33155733346939087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,balanced,0.34410667419433594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,balanced,0.367141326268514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.38517332077026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.43375468254089355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.4718986749649048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.5647573471069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.6526826620101929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.8408213456471761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,balanced,1.1073493162790935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,balanced,1.4647307395935059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,balanced,1.9555199940999348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.38360960483551027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.5737728118896485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,0.6451519966125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,balanced,0.07838933169841766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,balanced,0.07618666688601176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,balanced,0.07645866771539052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,balanced,0.07588799794514973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,balanced,0.07698133091131847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,balanced,0.07780266801516215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,balanced,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,balanced,0.08168533444404602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,balanced,0.08365866541862488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,balanced,0.08534399668375652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,balanced,0.08673600355784099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,balanced,0.08693333466847737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,balanced,0.09036800265312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,balanced,0.09798399607340495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,balanced,0.10450133681297302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,balanced,0.1179200013478597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,balanced,0.12498133381207784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,balanced,0.1495626668135325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,balanced,0.1906826694806417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,balanced,0.23921600977579752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,balanced,0.2844746708869934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.01,0.029216000437736513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.01,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.01,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.01,0.0735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.01,0.07665280103683472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.01,0.09727360010147094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.01,0.09870079755783082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.01,0.1056447982788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.01,0.11356799602508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.01,0.14090880155563354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.01,0.1404096007347107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.01,0.17424000501632692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.01,0.1995136022567749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.01,0.2596735954284668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.01,0.3124991893768311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.01,0.4640960216522217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.01,0.5349440097808837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,power_law_1.2,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,power_law_1.2,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,power_law_1.2,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,power_law_1.2,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,power_law_1.2,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,power_law_1.2,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.2,0.09091200232505799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.2,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.2,0.13296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.2,0.15224319696426392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.2,0.21583359241485595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.2,0.27331840991973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.2,0.3626368045806885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.2,0.45384960174560546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,power_law_1.01,0.03555200099945068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,power_law_1.01,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,power_law_1.01,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,power_law_1.01,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,power_law_1.01,0.05802879929542541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,power_law_1.01,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,power_law_1.01,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,power_law_1.01,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,power_law_1.01,0.166867196559906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,power_law_1.01,0.20378239154815675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,power_law_1.01,0.28695039749145507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,power_law_1.01,0.40067200660705565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.2,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.2,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.2,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.2,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.2,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.2,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.2,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.2,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.2,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.2,0.1270848035812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.2,0.16655999422073364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.2,0.1893504023551941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.2,0.2650496006011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.2,0.2742271900177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.2,0.3630592107772827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.2,0.45523838996887206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.2,0.6146624088287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.2,0.8386560440063476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.09156479835510253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.09134719967842102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.09058560132980346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.09591040015220642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.09624320268630981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.0952127993106842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.09240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.11751680374145508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.15635199546813966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.1673151969909668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.20396161079406738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.2618304014205933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.3389695882797241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.41607041358947755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,power_law_1.01,0.6230527877807617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,power_law_1.01,0.8035776138305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.01,0.18817919492721558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.01,0.1657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.01,0.1863487958908081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.01,0.2334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.01,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.01,0.23581440448760987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.01,0.23138558864593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.01,0.23991041183471679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.01,0.23824639320373536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.01,0.23712639808654784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.01,0.24287359714508056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.01,0.2506239891052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.01,0.25128960609436035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.01,0.2608896017074585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.01,0.2689536094665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.01,0.298419189453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.01,0.33960959911346433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.01,0.33331201076507566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.01,0.40047359466552734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.01,0.47207040786743165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.01,0.5749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.01,0.5707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.01,0.7251711845397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.01,0.8730624198913575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.01,1.1978879928588868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.01,1.4258943557739259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,balanced,0.0503359983364741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,balanced,0.07400533556938171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,balanced,0.09021332859992981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,balanced,0.11742400129636128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,balanced,0.15562666455904642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,balanced,0.17332265774408975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,balanced,0.24316799640655518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,balanced,0.30080000559488934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,balanced,0.434005339940389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,balanced,0.5611573457717896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,balanced,0.05900266766548157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,balanced,0.056736002365748085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,balanced,0.06599999964237213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,balanced,0.06857599814732869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,balanced,0.08955732981363933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,balanced,0.10275200009346008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,balanced,0.12549866239229837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,balanced,0.14455999930699667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,balanced,0.19754666090011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,balanced,0.22885332504908243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,balanced,0.3155786593755086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,balanced,0.3924320141474406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,power_law_1.2,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,power_law_1.2,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,power_law_1.2,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,power_law_1.2,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,power_law_1.2,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,power_law_1.2,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,power_law_1.2,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,power_law_1.2,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,power_law_1.2,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,power_law_1.2,0.13073279857635497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,power_law_1.2,0.1472383975982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,power_law_1.2,0.17358720302581787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.2,0.21437439918518067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.2,0.32268800735473635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.2,0.3973504066467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.2,0.5800447940826416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.2,0.759929609298706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.2,1.064799976348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,power_law_1.01,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,power_law_1.01,0.028575998544692994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,power_law_1.01,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,power_law_1.01,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,power_law_1.01,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,power_law_1.01,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,power_law_1.01,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,power_law_1.01,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,power_law_1.01,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,power_law_1.01,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,power_law_1.01,0.08861439824104309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,power_law_1.01,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.2,1.4774335861206054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,power_law_1.01,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,power_law_1.01,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,power_law_1.01,0.09127680063247681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,power_law_1.01,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,power_law_1.01,0.09390720129013061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,power_law_1.01,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.01,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.01,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,power_law_1.01,0.11159039735794067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.01,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.01,0.1564095973968506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.01,0.2167423963546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.01,0.2740288019180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,power_law_1.01,0.11119359731674194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.01,0.44301438331604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,power_law_1.01,0.11383039951324463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.01,0.5919360160827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,power_law_1.01,0.1326848030090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.01,0.1491968035697937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.01,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.01,0.2004096031188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.01,0.26615679264068604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.01,0.32027521133422854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.01,0.454860782623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.01,0.5935359954833984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.01,0.8755200386047364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.01,1.0925824165344238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.2,2.1959487915039064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.2,2.8446975708007813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,balanced,0.041189332803090416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,balanced,0.0401706670721372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,balanced,0.04268266757329305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,balanced,0.07868800063927968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,balanced,0.0806826651096344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,balanced,0.08144533137480418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,balanced,0.08221333225568135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,balanced,0.0846560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,balanced,0.08653333783149719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,balanced,0.09032000104586284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,balanced,0.09156800309816997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,balanced,0.1030613382657369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,balanced,0.10629333058993022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,balanced,0.11954133709271748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,balanced,0.13358400265375772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,balanced,0.1593226691087087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,balanced,0.18149334192276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,balanced,0.23873066902160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,balanced,0.28253332773844403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,balanced,0.39636798699696857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,balanced,0.4920586744944255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,power_law_1.2,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,power_law_1.2,0.021990400552749634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,power_law_1.2,0.02311040014028549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,power_law_1.2,0.02369280010461807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,power_law_1.2,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,power_law_1.2,0.02632960081100464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,power_law_1.2,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,power_law_1.2,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,power_law_1.2,0.04048640131950378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,power_law_1.2,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.09526399970054626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.20561280250549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,power_law_1.01,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,power_law_1.01,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,power_law_1.01,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,power_law_1.01,0.05764480233192444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,power_law_1.01,0.0695743978023529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,power_law_1.01,0.12663040161132813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,power_law_1.01,0.1621567964553833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,power_law_1.01,0.2187135934829712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,power_law_1.01,0.28834559917449953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,power_law_1.01,0.36450560092926027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,balanced,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,balanced,0.09005332986513774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,balanced,0.09206400314966838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,balanced,0.0943999985853831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,balanced,0.09293333689371745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,balanced,0.09360532959302266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,balanced,0.09431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,balanced,0.09758399923642476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,balanced,0.097871998945872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,balanced,0.11601066589355469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,balanced,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,balanced,0.12034133076667786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,balanced,0.142384002606074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.16451733311017355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.20297600825627646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.22184000412623087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.28542399406433105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.3555146853129069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.49565335114796955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.6321066617965698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,balanced,0.91267196337382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,balanced,1.1953226725260417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.28108160495758056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,power_law_1.2,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,power_law_1.2,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,power_law_1.2,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,power_law_1.2,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,power_law_1.2,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,power_law_1.2,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,power_law_1.2,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,power_law_1.2,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,power_law_1.2,0.0878271996974945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,power_law_1.2,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,power_law_1.2,0.10120320320129395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,power_law_1.2,0.11078399419784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,power_law_1.2,0.12947839498519897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,power_law_1.2,0.1559424042701721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,power_law_1.2,0.16310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,power_law_1.2,0.20115199089050292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,power_law_1.2,0.23252480030059813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,power_law_1.2,0.31485440731048586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,power_law_1.2,0.4024831771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,power_law_1.2,0.5717887878417969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,power_law_1.2,0.7765632152557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,balanced,0.07147199908892314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,balanced,0.0993386705716451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,balanced,0.15664533774058023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,balanced,0.2641493280728658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,balanced,0.26683199405670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,balanced,0.2717439929644267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,balanced,0.27105599641799927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,balanced,0.2739680012067159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,balanced,0.27793065706888836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,balanced,0.27818665901819867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,balanced,0.28412266572316486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,balanced,0.2871840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,balanced,0.2921813329060872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,balanced,0.301744004090627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,balanced,0.30899200836817425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,balanced,0.32547199726104736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.06747519969940186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.1289728045463562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,balanced,0.3573066790898641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.3955306609471639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.16468479633331298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.45132267475128174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.5056053400039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.19617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.7048160235087076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.2906496047973633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.784874677658081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,balanced,1.127679983774821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.37166719436645507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.3690346082051594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.48113279342651366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,balanced,1.9766453107198079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.5932223796844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,balanced,2.6069653828938804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,power_law_1.2,0.043347200751304625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,power_law_1.2,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,power_law_1.2,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,power_law_1.2,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,power_law_1.2,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,power_law_1.2,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,power_law_1.2,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,power_law_1.2,0.07051519751548767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.2,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.2,0.10012160539627075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.2,0.11627520322799682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.2,0.1493183970451355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.2,0.18260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.2,0.25489919185638427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.2,0.34837119579315184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.2,0.5154047966003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.2,0.6872128009796142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,power_law_1.2,0.09821439981460571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,power_law_1.2,0.14659199714660645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,power_law_1.2,0.2153791904449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,power_law_1.2,0.3019968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,power_law_1.2,0.41947522163391116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,power_law_1.2,0.5701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,power_law_1.2,0.6733759880065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,power_law_1.2,0.8647871971130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,power_law_1.2,0.8913727760314941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,power_law_1.2,0.9194944381713868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,power_law_1.2,0.9941503524780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,power_law_1.2,0.9334848403930665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,power_law_1.2,0.9197504043579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,power_law_1.2,0.9995648384094238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,power_law_1.2,1.0508288383483886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,power_law_1.2,1.0847552299499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,power_law_1.2,1.1140416145324707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,power_law_1.2,1.2485568046569824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,power_law_1.2,1.2714495658874512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,power_law_1.2,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,power_law_1.2,1.4194304466247558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,power_law_1.2,0.06344959735870362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,power_law_1.2,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,power_law_1.2,1.5250816345214844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,power_law_1.2,1.8087871551513672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,power_law_1.2,0.0821183979511261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,power_law_1.2,1.9643583297729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,power_law_1.2,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,power_law_1.2,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,power_law_1.2,2.482988739013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,power_law_1.2,0.09034240245819092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,power_law_1.2,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,power_law_1.2,0.09853439927101135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,power_law_1.2,2.6029056549072265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,power_law_1.2,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,power_law_1.2,0.10955519676208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,power_law_1.2,0.12053760290145873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.2,0.13163520097732545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,power_law_1.2,3.3971710205078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.2,0.1592960000038147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.2,0.18108799457550048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.2,0.2371392011642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.2,0.2834752082824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,power_law_1.2,4.140006256103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.2,0.36689279079437254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.2,0.5373951911926269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.2,0.7812479972839356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.2,0.9562687873840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,balanced,0.05026133358478546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,balanced,0.05851200222969055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,balanced,0.060271998246510826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,balanced,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,balanced,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,balanced,0.06806399921576183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,balanced,0.06832000116507213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,balanced,0.07020799815654755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,balanced,0.07530133426189423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,balanced,0.08458133538564046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,balanced,0.09353066484133403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,balanced,0.1053706705570221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,balanced,0.13079999883969626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,balanced,0.1507253348827362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,balanced,0.19205333789189658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,balanced,0.22966933250427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,balanced,0.3078026572863261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,balanced,0.3895786603291829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,balanced,0.5468480189641317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,balanced,0.703167994817098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,power_law_1.2,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,power_law_1.2,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,power_law_1.2,0.08637440204620361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,power_law_1.2,0.08398720026016235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,power_law_1.2,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,power_law_1.2,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,power_law_1.2,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,power_law_1.2,0.0902463972568512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,power_law_1.2,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,power_law_1.2,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,power_law_1.2,0.10669440031051636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,power_law_1.2,0.11362559795379638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,power_law_1.2,0.12653440237045288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.2,0.13434879779815673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.2,0.18036479949951173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.2,0.19708800315856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.2,0.2608448028564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.2,0.3193664073944092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.2,0.4543168067932129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.2,0.6224127769470215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.2,1.0241920471191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.2,1.239635181427002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,balanced,0.05363733569780985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,balanced,0.06939200063546498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,balanced,0.1467413306236267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,balanced,0.24506133794784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,balanced,0.3386666774749756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,balanced,0.34329601128896076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,balanced,0.34515198071797687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,balanced,0.34519465764363605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,balanced,0.3500906626383464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,balanced,0.350874662399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,balanced,0.35126932462056476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,balanced,0.348746657371521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,balanced,0.3556533257166545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,balanced,0.35533865292867023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,balanced,0.36188264687856037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,balanced,0.37323200702667236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,balanced,0.3763573169708252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,balanced,0.3930773337682088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,balanced,0.41095467408498126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,balanced,0.4427573283513387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,balanced,0.47997331619262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,balanced,0.5366400082906088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,balanced,0.6351573467254639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,balanced,0.744821310043335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,balanced,0.9504906336466471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,balanced,0.064560001095136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,balanced,0.09245866537094116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,balanced,0.1390773355960846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,balanced,0.14220800002415976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,balanced,0.14291733503341675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,balanced,0.14453333616256714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,balanced,0.14619732896486917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,balanced,0.14669332901636759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,balanced,0.14677332838376364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,balanced,0.14839999874432883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,balanced,0.14973333477973938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,balanced,0.15194132924079895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,balanced,0.15215466419855753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,balanced,0.1569439967473348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,balanced,0.17521067460378012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,balanced,0.17644800742467245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,balanced,0.18281600872675577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,balanced,0.22028799851735434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.23645333449045816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.2909280061721802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.35468800862630206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.4735413392384847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.5685386657714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.8042506376902262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.0300959746042888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.4868639310201008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,balanced,1.9561653137207031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,balanced,0.07576000193754832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,balanced,0.07799466451009114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,balanced,0.07678399980068207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,balanced,0.07839466631412506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,balanced,0.07838933169841766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,balanced,0.08633599678675334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,balanced,0.0867680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,balanced,0.09128000338872273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,balanced,0.09748799602190654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,balanced,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,balanced,0.11114133397738139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,balanced,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,balanced,0.13260799646377563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,balanced,0.16482133666674295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,balanced,0.1858773430188497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,balanced,0.25443732738494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,balanced,0.3110613425572713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,balanced,0.43038400014241535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,balanced,0.5697493155797323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,power_law_1.2,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,power_law_1.2,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,power_law_1.2,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,power_law_1.2,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,power_law_1.2,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,power_law_1.2,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,power_law_1.2,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.2,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.2,0.137881600856781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.2,0.17671040296554566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.1400063991546631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.2,0.21671040058135987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.2331775903701782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.2,0.2854464054107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.21656320095062256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.2,0.3693376064300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.23866240978240966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.2,0.5134975910186768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.23767681121826173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.2,0.7085887908935546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.21595520973205568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.2133568048477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.21628160476684571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.2204416036605835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.22587521076202394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.2158207893371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,balanced,0.1202826698621114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.22919681072235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,balanced,0.2113866607348124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,balanced,0.21545066436131796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,balanced,0.21035200357437134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.21792640686035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,balanced,0.2087093393007914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,balanced,0.20874667167663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,balanced,0.20918933550516763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.22895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,balanced,0.21042132377624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,balanced,0.20805333058039346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,balanced,0.21846934159596762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.2624255895614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,balanced,0.21236799160639444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,balanced,0.21305066347122192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,balanced,0.23310933510462442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.263372802734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,balanced,0.23931733767191568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,balanced,0.24266666173934937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.2744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.3129472017288208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.31208319664001466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.36469759941101076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.38555519580841063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,balanced,0.2505653301874797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.44867839813232424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,balanced,0.2890453338623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,balanced,0.29611732562383014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,balanced,0.37094398339589435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.5230976104736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,balanced,0.41857067743937176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,balanced,0.5603733460108439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,balanced,0.6934186617533366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,balanced,0.9661760330200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,balanced,1.2539520263671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,balanced,1.8517492612202961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,balanced,2.444767951965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,power_law_1.2,0.6880127906799316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,power_law_1.2,0.8563520431518554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.1681856155395507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,power_law_1.2,1.490886402130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.14382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.2658368110656738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.32951040267944337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.4192255973815918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.3870464086532593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.5308032035827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.5820159912109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.5833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.5153600215911865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.5439424037933349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.5821119785308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.5755263805389405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.5910336017608643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.5293439865112305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.5820928096771241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.6209919929504395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.681990385055542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,power_law_1.01,0.7743360042572022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,power_law_1.01,0.7443456172943115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,power_law_1.01,0.8322815895080566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,power_law_1.01,0.9959360122680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,power_law_1.01,1.2178688049316406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,power_law_1.01,1.2908991813659667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,power_law_1.01,1.6156991958618163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,power_law_1.01,2.106553649902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,power_law_1.2,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,power_law_1.2,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,power_law_1.2,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,power_law_1.2,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,power_law_1.2,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.2,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.2,0.10732799768447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.2,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.2,0.16455680131912231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.2,0.20689280033111573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.2,0.2831295967102051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.2,0.3676352024078369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.2,0.5907008171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.2,0.8415936470031739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,1,balanced,0.159578671058019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,1,balanced,0.1650986671447754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,1,balanced,0.17466133832931519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,1,balanced,0.2064746618270874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,1,balanced,0.2730986674626668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,1,balanced,0.42344534397125244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,1,balanced,0.4238400061925252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,1,balanced,0.4309440056482951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,1,balanced,0.4304106632868449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,1,balanced,0.43513067563374835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,1,balanced,0.43644265333811444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,1,balanced,0.44312000274658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,1,balanced,0.4455039898554484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,1,balanced,0.45342934131622314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,1,balanced,0.4506186644236247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,1,balanced,0.4551946719487508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,1,balanced,0.4662026564280192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,1,balanced,0.48974935213724774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,1,balanced,0.5176639954249064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,1,balanced,0.5796693166097006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,1,balanced,0.6413333415985107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,1,balanced,0.7309066454569498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,1,balanced,0.8401813507080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,1,balanced,1.180239995320638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,1,balanced,1.4077547391255696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,1,balanced,1.9926133155822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,1,balanced,2.6267199516296387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,balanced,0.03992533435424169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,balanced,0.059808000922203064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,balanced,0.06244266529877981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,balanced,0.06554666658242543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,balanced,0.06666133304437001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,balanced,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,balanced,0.08741333087285359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,balanced,0.09773866335550944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,balanced,0.12484799822171529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,balanced,0.16245866815249124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,balanced,0.2026080091794332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,balanced,0.24138667186101279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,balanced,0.33504001299540204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,balanced,0.42265601952870685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,balanced,0.6122826735178629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,balanced,0.7843946615854899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,power_law_1.01,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,power_law_1.01,0.1420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,power_law_1.01,0.1678655982017517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,power_law_1.01,0.19339519739151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,power_law_1.01,0.21086080074310304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,power_law_1.01,0.2164544105529785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,power_law_1.01,0.241593599319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,power_law_1.01,0.2144063949584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,power_law_1.01,0.20899200439453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,power_law_1.01,0.2118272066116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,power_law_1.01,0.22577919960021972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,power_law_1.01,0.21480960845947267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,power_law_1.01,0.23158400058746337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,power_law_1.01,0.2443392038345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,power_law_1.01,0.23982079029083253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,power_law_1.01,0.24183039665222167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,power_law_1.01,0.26182401180267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,power_law_1.01,0.3071808099746704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,power_law_1.01,0.3080064058303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,power_law_1.01,0.35386879444122316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,power_law_1.01,0.3675584077835083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,power_law_1.01,0.44736638069152834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,power_law_1.01,0.5371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,power_law_1.01,0.6915008068084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,power_law_1.01,0.8410367965698242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,power_law_1.01,1.1182592391967774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,power_law_1.01,1.439891242980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.0626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,balanced,0.061039999127388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,balanced,0.06551466882228851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,balanced,0.07076799869537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,balanced,0.09217600027720134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,balanced,0.09307733178138733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,balanced,0.09724799791971843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,balanced,0.09917866190274556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,balanced,0.0974666674931844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,balanced,0.0974773367245992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,balanced,0.09920000036557515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,balanced,0.10860266288121541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,balanced,0.11272000273068745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,balanced,0.11453333497047424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,balanced,0.1125386655330658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,balanced,0.11930132905642192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,balanced,0.12989866733551025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,balanced,0.150709331035614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,balanced,0.16993600130081177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,balanced,0.21194666624069214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,balanced,0.2481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,balanced,0.34784531593322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,balanced,0.4247200091679891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.12592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.13744640350341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.18697600364685057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.23286399841308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.33223040103912355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.4465792179107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,balanced,0.07208533088366191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,balanced,0.0988159974416097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,balanced,0.1276586651802063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,balanced,0.12423466642697652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,balanced,0.12571733196576437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,balanced,0.12381333112716675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,balanced,0.1288533310095469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,balanced,0.129013329744339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,balanced,0.12931199868520102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,balanced,0.1295413374900818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,balanced,0.13833066821098328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,balanced,0.13596799969673157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,balanced,0.1414186656475067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,balanced,0.14918399850527445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,balanced,0.154639999071757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,balanced,0.1692426602045695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,balanced,0.1819146672884623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,balanced,0.21480000019073486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,balanced,0.24532800912857056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,balanced,0.3030293385187785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,balanced,0.3684426546096802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,balanced,0.4835413297017415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,balanced,0.6058346827824911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,power_law_1.2,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,power_law_1.2,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,power_law_1.2,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,power_law_1.2,0.07822080254554749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,power_law_1.2,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,power_law_1.2,0.0871295988559723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,power_law_1.2,0.0974016010761261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,power_law_1.2,0.09749119877815246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,power_law_1.2,0.11690239906311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.2,0.12639360427856444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.2,0.1529728055000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.2,0.17797759771347046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.2,0.24753279685974122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.2,0.3015615940093994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.2,0.4681407928466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.2,0.6001152038574219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.2,0.9003904342651368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.2,1.3476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.0719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.10764800310134888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.1531008005142212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.15543680191040038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.1652672052383423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.1596351981163025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.166975998878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.17167999744415283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.17474559545516968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.18072960376739503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.19467519521713256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.19830399751663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.2088383913040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.23400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.2627968072891235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.30976641178131104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.3475712060928345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.4651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.556550407409668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.7774975776672364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.941260814666748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,1.3578175544738769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,1.8262271881103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.01,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.01,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.01,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.01,0.0759935975074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.01,0.07171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.01,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.01,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.01,0.08264960050582885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.01,0.0877120018005371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.01,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.01,0.096806401014328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.01,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.01,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.01,0.12236160039901733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.01,0.1371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.01,0.16672639846801757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.01,0.18804479837417604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.01,0.23717761039733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.01,0.3035072088241577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.01,0.3966655969619751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.01,0.48638081550598145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,balanced,0.05133333305517832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,balanced,0.04819199939568838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,balanced,0.053690666953722634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,balanced,0.0543146679798762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,balanced,0.05457599957784017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,balanced,0.05659733215967814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,balanced,0.0644160012404124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,balanced,0.06839466591676076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,balanced,0.07081066568692525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,balanced,0.08456533153851827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,balanced,0.09685867031415303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,balanced,0.11943466464678447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,balanced,0.13593066732088724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,balanced,0.1790613333384196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,balanced,0.20152533054351807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,balanced,0.27610133091608685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,balanced,0.344650665918986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.01,0.02871040105819702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.01,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.01,0.029862400889396668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.01,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.01,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.01,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.01,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.01,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.01,0.11369600296020507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.01,0.1476096034049988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.01,0.1894271969795227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.01,0.2262592077255249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.2,0.027846398949623107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.2,0.03155840039253235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.2,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.2,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.2,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.2,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.2,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.2,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.2,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.2,0.15853439569473265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.2,0.16047359704971315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.2,0.24947199821472169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.2,0.24859519004821778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,balanced,0.06622399886449178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,balanced,0.08319999774297078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,balanced,0.11153067151705424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,balanced,0.1655946671962738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,balanced,0.25086400906244916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,balanced,0.3250453273455302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,balanced,0.32662399609883624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,balanced,0.3311786651611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,balanced,0.331221342086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,balanced,0.33477866649627686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,balanced,0.33881068229675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,balanced,0.3433599869410197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,balanced,0.34485332171122235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,balanced,0.3519253333409627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,balanced,0.3596479892730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,balanced,0.3641386826833089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,balanced,0.3776693344116211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,balanced,0.4037919839223226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,balanced,0.4256266752878825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,balanced,0.5586400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,balanced,0.5454080104827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,balanced,0.7939573129018148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,balanced,0.7889653046925863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,balanced,1.149994691212972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,balanced,1.2887946764628093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,balanced,1.9229386647542317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,balanced,2.2762932777404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,power_law_1.01,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,power_law_1.01,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,power_law_1.01,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,power_law_1.01,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,power_law_1.01,0.10737919807434082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,power_law_1.01,0.10892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,power_law_1.01,0.10967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,power_law_1.01,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,power_law_1.01,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,power_law_1.01,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,power_law_1.01,0.12344959974288941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,power_law_1.01,0.13891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,power_law_1.01,0.14540799856185913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,power_law_1.01,0.16290559768676757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,power_law_1.01,0.18823039531707764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,power_law_1.01,0.22581119537353517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,power_law_1.01,0.27100799083709715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,power_law_1.01,0.34847359657287597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,power_law_1.01,0.4214911937713623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,power_law_1.01,0.5652671813964844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,power_law_1.01,0.7273727893829346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,power_law_1.2,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,power_law_1.2,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,power_law_1.2,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,power_law_1.2,0.0940671980381012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,power_law_1.2,0.10091520547866821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,power_law_1.2,0.10243840217590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,power_law_1.2,0.10392960309982299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,power_law_1.2,0.11083519458770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,power_law_1.2,0.1185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,power_law_1.2,0.12171519994735717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,power_law_1.2,0.13633919954299928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.2,0.1505728006362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.2,0.17214720249176024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.2,0.19719040393829346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.2,0.25004799365997316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.2,0.3049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.2,0.39784319400787355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.2,0.5116032123565674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.2,0.7169536113739013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.2,0.9526464462280273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,power_law_1.2,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,power_law_1.2,0.0282368004322052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,power_law_1.2,0.028198400139808656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,power_law_1.2,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,power_law_1.2,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,power_law_1.2,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,power_law_1.2,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,power_law_1.2,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,power_law_1.2,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,power_law_1.2,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,power_law_1.2,0.09871360063552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,power_law_1.2,0.1404096007347107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,power_law_1.2,0.16640000343322753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,power_law_1.2,0.23544960021972655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,power_law_1.2,0.2750272035598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.10673919916152955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.10734080076217652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.11201920509338378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.11396479606628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.11021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.1172287940979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.11800960302352906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.12698240280151368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.1329856038093567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.14230400323867798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.16826879978179932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.2079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.1965183973312378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.25628159046173093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.26472959518432615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,power_law_1.01,0.3510080099105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,power_law_1.01,0.42030720710754393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,power_law_1.01,0.5747200012207031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,power_law_1.01,0.8096128463745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,power_law_1.2,0.3918463945388794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,power_law_1.01,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,power_law_1.01,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,power_law_1.01,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,power_law_1.01,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,power_law_1.2,0.5236671924591064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,power_law_1.01,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,power_law_1.01,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,power_law_1.01,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,power_law_1.01,0.08216320276260376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,power_law_1.01,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,power_law_1.01,0.08382080197334289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,power_law_1.01,0.09200000166893005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,power_law_1.01,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,power_law_1.01,0.09819520115852357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,power_law_1.01,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.01,0.11474560499191284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.01,0.13764480352401734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.01,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.01,0.20776960849761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.01,0.23740160465240479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.01,0.372761607170105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.01,0.49259519577026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.01,0.7530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.01,0.9663295745849609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,balanced,0.084714670976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,balanced,0.10875200231870015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,balanced,0.15445866187413534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,balanced,0.23544534047444662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,balanced,0.3903679847717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,balanced,0.3964746793111165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,balanced,0.4033120075861613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,balanced,0.40933334827423096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,balanced,0.41047998269399005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,balanced,0.40778132279713947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,balanced,0.4165866772333781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,balanced,0.4148586591084798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,balanced,0.421776016553243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,balanced,0.4290879964828491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,balanced,0.4331466754277547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,balanced,0.4472959836324056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,balanced,0.49006934960683185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.5194986661275228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.5975199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.6597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.8132373491923014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.9657866954803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,balanced,1.3778293927510579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,balanced,1.665445327758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,balanced,2.4286559422810874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,balanced,3.1492319107055664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.07707520127296448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.11662080287933349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.11671680212020874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.13068799972534179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.13329919576644897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.13872640132904052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.14488960504531861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.14320640563964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.14376319646835328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.14974720478057862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.15429120063781737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.16499199867248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.1811840057373047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.19768320322036742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.23303680419921874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.263923192024231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.3344127893447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.4011839866638184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.5421440124511718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.6829055786132813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,0.9760000228881835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.2271295547485352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.06771199703216553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.09041280150413514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.09351040124893188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.09675520062446594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.10357760190963745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.11918720006942748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.13891199827194214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.14818559885025023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.17843199968338014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.2212735891342163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.2720896005630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.31251840591430663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.42616958618164064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.5581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.7971583843231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,1.0036992073059081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,power_law_1.01,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,power_law_1.01,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,power_law_1.01,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,power_law_1.01,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,power_law_1.01,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,power_law_1.01,0.08602240085601806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,power_law_1.01,0.10527360439300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,power_law_1.01,0.1251904010772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,power_law_1.01,0.1440384030342102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,power_law_1.01,0.17637759447097778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,power_law_1.01,0.21636478900909423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,power_law_1.01,0.29047040939331054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,power_law_1.01,0.3669312000274658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,power_law_1.01,0.5240128040313721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,power_law_1.01,0.6532351970672607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,power_law_1.01,0.10506240129470826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,power_law_1.01,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,power_law_1.01,0.1297727942466736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,power_law_1.01,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,power_law_1.01,0.1804479956626892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,power_law_1.01,0.18642560243606568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,power_law_1.01,0.19393919706344603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,power_law_1.01,0.20478720664978028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,power_law_1.01,0.20343680381774903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,power_law_1.01,0.20019838809967042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,power_law_1.01,0.2106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,power_law_1.01,0.20107519626617432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,power_law_1.01,0.21704320907592772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,power_law_1.01,0.22154879570007324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,4,power_law_1.01,0.22018558979034425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,4,power_law_1.01,0.2273024082183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,4,power_law_1.01,0.2384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.01,0.2539776086807251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.01,0.28108160495758056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.01,0.31036159992218015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.01,0.38439040184020995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.01,0.44851198196411135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.01,0.6024831771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.01,0.7575615882873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.01,1.0699584007263183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.01,1.3892352104187011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.11642240285873413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.14895999431610107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.20055038928985597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.3220031976699829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.4731584072113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.6613120079040528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.7431488037109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.9993727684020997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,1.0685055732727051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,1.0717568397521973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,1.151046371459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,1.1773568153381349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,1.2407679557800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,1.280684757232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,1.278054428100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,1.3077119827270507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,1.3873536109924316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,1.4711423873901368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.5492863655090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.7334207534790038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,1.942086410522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,2.2767168045043946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,2.4464767456054686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,3.0009344100952147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.2,3.5880897521972654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.2,4.731520080566407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.2,5.8760833740234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,power_law_1.2,0.09062399864196777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,power_law_1.2,0.1411072015762329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,power_law_1.2,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,power_law_1.2,0.16743680238723754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,power_law_1.2,0.20176000595092775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,power_law_1.2,0.2248447895050049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,power_law_1.2,0.21332480907440185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,power_law_1.2,0.21406080722808837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,power_law_1.2,0.2286463975906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,balanced,0.07900266846021016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,power_law_1.2,0.2198784112930298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,balanced,0.07918400069077809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,balanced,0.08117866516113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,power_law_1.2,0.23303039073944093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,power_law_1.2,0.22611839771270753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,power_law_1.2,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,power_law_1.2,0.24277760982513427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,power_law_1.2,0.24603519439697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,power_law_1.2,0.25557119846343995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,balanced,0.08301333089669545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,balanced,0.084197332461675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,balanced,0.08710400263468425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,balanced,0.09318400422732036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,balanced,0.09206400314966838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,balanced,0.09710400303204854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,balanced,0.10547199845314026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,balanced,0.10974400242169698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,balanced,0.12500266234079996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,balanced,0.1525173286596934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,balanced,0.18001067638397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,balanced,0.2085813283920288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,balanced,0.27217066287994385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,balanced,0.33685866991678876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,balanced,0.4617439905802409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,balanced,0.5837920109430949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,power_law_1.2,0.27331840991973877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,power_law_1.2,0.32537600994110105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,power_law_1.2,0.31699841022491454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,power_law_1.2,0.36894080638885496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,power_law_1.2,0.39724159240722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,power_law_1.2,0.4940032005310059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,power_law_1.2,0.5636223793029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,power_law_1.2,0.7276864051818848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,power_law_1.2,0.9386048316955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,power_law_1.2,1.2816127777099608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,power_law_1.2,1.6883199691772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.08474239706993103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.14713599681854247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.14913920164108277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.1491711974143982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.15143680572509766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.15147520303726197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.1526528000831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.15555200576782227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.1597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.16881279945373534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.19228800535202026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.19575040340423583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.217958402633667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.2517375946044922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.3094208002090454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.3860608100891113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.4379583835601807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.6430784225463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,balanced,0.04717866579691569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,balanced,0.06039999922116598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,balanced,0.06146133442719778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,balanced,0.0620000014702479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,balanced,0.06512533128261566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,balanced,0.0728000005086263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,balanced,0.09413333733876546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,balanced,0.13170666495958963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,balanced,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,balanced,0.20444266001383463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,balanced,0.19557867447535196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,balanced,0.2601173321406047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,balanced,0.3073386748631795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,balanced,0.4209333260854085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,balanced,0.5142560005187988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.7094272136688232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,power_law_1.01,0.07229440212249756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,power_law_1.01,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,power_law_1.01,0.09120640158653259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,power_law_1.01,0.10874240398406983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,power_law_1.01,0.11138559579849243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,power_law_1.01,0.1145408034324646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,power_law_1.01,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,power_law_1.01,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.168518352508545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,power_law_1.01,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,power_law_1.01,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,power_law_1.01,0.12525440454483033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,power_law_1.01,0.1277567982673645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,power_law_1.01,0.12863999605178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,power_law_1.01,0.13284480571746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,power_law_1.01,0.14046080112457277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,power_law_1.01,0.16562559604644775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.01,0.18625919818878173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.01,0.21875200271606446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.01,0.2724031925201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.01,0.3651263952255249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.01,0.45815038681030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.01,0.6770431995391846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.01,0.9193856239318847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.3079872131347656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.01,1.2213312149047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.01,1.6348991394042969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.12462079524993896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.13777920007705688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.140665602684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.13923840522766112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.1229632019996643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.13415679931640626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.13715840578079225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.13351680040359498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.14186240434646608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.14453760385513306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.1579200029373169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.18929920196533204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.21644160747528077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.23837440013885497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,2.1167743682861326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.34124159812927246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.35260159969329835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.478220796585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.5704959869384766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,0.8015680313110352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,1.1167807579040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,2.8706560134887695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,balanced,0.03649600098530451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,balanced,0.07707199951012929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,balanced,0.124208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,balanced,0.1253493328889211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,balanced,0.12717866897583008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,balanced,0.1272160013516744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,balanced,0.12763200203577676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,balanced,0.12873066465059915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,balanced,0.1304373343785604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,balanced,0.13179199894269308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,balanced,0.13821333646774292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,balanced,0.14083733161290488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,balanced,0.14653866489728293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,balanced,0.1524799962838491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,balanced,0.17124267419179282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,balanced,0.1813919941584269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,balanced,0.21874133745829263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,balanced,0.2387253244717916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,balanced,0.3118826746940613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,balanced,0.36177066961924237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,balanced,0.4984533389409383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,balanced,0.6178826491038004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,balanced,0.8731839656829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,balanced,1.1537333329518635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,balanced,0.03384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,balanced,0.03626666714747747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,balanced,0.036176001032193504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,balanced,0.04260266820589701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.08362133304278056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.09577066699663798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.12310399611790974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.1463573376337687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.06305919885635376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.0657472014427185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.06443520188331604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.06723840236663818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.11706880331039429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.1372607946395874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.1746240019798279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.2058624029159546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.2870847940444946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.35237119197845457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.49976320266723634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.6419136047363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.01,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.01,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.01,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.01,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.01,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.01,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.01,0.042483198642730716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.01,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.01,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.01,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.01,0.10684159994125367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.01,0.12607359886169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.01,0.1665536046028137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.01,0.2107840061187744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,power_law_1.01,0.06615679860115051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,power_law_1.01,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,power_law_1.01,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,power_law_1.01,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,power_law_1.01,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,power_law_1.01,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,power_law_1.01,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,power_law_1.01,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,power_law_1.01,0.09615359902381897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,power_law_1.01,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,power_law_1.01,0.09861760139465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,power_law_1.01,0.1031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,power_law_1.01,0.10769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,power_law_1.01,0.10973440408706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,power_law_1.01,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.01,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.01,0.14911999702453613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.01,0.17464959621429443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.01,0.21273601055145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.01,0.24279680252075195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.01,0.32848639488220216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.01,0.41441922187805175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.01,0.5562111854553222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.01,0.7423871994018555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.10931199789047241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.13534079790115355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.1390079975128174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.13841919898986815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.13805439472198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.14728959798812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.14415359497070312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.15314559936523436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.1617408037185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.16955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.17656960487365722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.17648639678955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.19751039743423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.20816640853881835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.2410048007965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.2816319942474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.3442431926727295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.41905918121337893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,power_law_1.01,0.5319744110107422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,power_law_1.01,0.6922880172729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,power_law_1.01,0.9195136070251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,power_law_1.01,1.1944128036499024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.2,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.2,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.2,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.2,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.2,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.2,0.04232960045337677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.2,0.04378879964351654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.2,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.2,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.2,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.2,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.2,0.14825600385665894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.2,0.2269887924194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.2,0.2785088062286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,balanced,0.051669334371884666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,balanced,0.07368533313274384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,balanced,0.08924800157546997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,balanced,0.08958400289217631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,balanced,0.09041066964467366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,balanced,0.09036266803741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,balanced,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,balanced,0.09237333138783772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,balanced,0.09176533420880635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,balanced,0.09139733513196309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,balanced,0.09954133629798889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,balanced,0.09857066472371419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,balanced,0.10341866811116536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,balanced,0.10981866717338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,balanced,0.11412800351778667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,balanced,0.1286079982916514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,balanced,0.1358026663462321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,balanced,0.16085867087046304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,balanced,0.18798933426539102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,balanced,0.2356320023536682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,balanced,0.28739200035731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,balanced,0.38069331645965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,balanced,0.47997868061065674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.08757759928703308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.08657919764518737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.0909056007862091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.09330559968948364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.09533439874649048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.0967423975467682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.10542080402374268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.1128767967224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.13626240491867064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.14318079948425294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.1749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.2024384021759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.25038719177246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.3087807893753052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.4077951908111572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.5454592227935791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.6955008029937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,0.8737792015075684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,power_law_1.01,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,power_law_1.01,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,power_law_1.01,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,power_law_1.01,0.06795520186424256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,power_law_1.01,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,power_law_1.01,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,power_law_1.01,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,power_law_1.01,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,power_law_1.01,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,power_law_1.01,0.09450240135192871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.01,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.01,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.01,0.14576640129089355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.01,0.21174399852752684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.01,0.2569216012954712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.01,0.40417919158935545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.01,0.5229375839233399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.01,0.6822271823883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.01,0.8984512329101563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.10151679515838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.10389759540557861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.10711040496826171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.1091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.11162240505218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.11322879791259766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.11758079528808593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.12334719896316529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.12982399463653566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.13559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.17601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.17187199592590333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.21103999614715577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.2331455945968628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.2945472002029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,power_law_1.2,0.38068480491638185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,power_law_1.2,0.5097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,power_law_1.2,0.5505407810211181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,power_law_1.2,0.863315200805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,power_law_1.2,0.9254336357116699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.07562239766120911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.08622080087661743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.09221760034561158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.13397120237350463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.14933760166168214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.174399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.24555521011352538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.28587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,power_law_1.2,0.3933887958526611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,power_law_1.2,0.5560704231262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.1508288025856018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.19285119771957399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.2701375961303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,power_law_1.01,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,power_law_1.01,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,power_law_1.01,0.08162559866905213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,power_law_1.01,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,power_law_1.01,0.12040319442749023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,power_law_1.01,0.15181440114974976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,power_law_1.01,0.16032639741897584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,power_law_1.01,0.16240639686584474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,power_law_1.01,0.16628479957580566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,power_law_1.01,0.17827199697494506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,power_law_1.01,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,power_law_1.01,0.1856063961982727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,power_law_1.01,0.2062079906463623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,power_law_1.01,0.2149888038635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,power_law_1.01,0.2223360061645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,power_law_1.01,0.21962239742279052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,power_law_1.01,0.2471872091293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,power_law_1.01,0.2692863941192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,power_law_1.01,0.32572159767150877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,power_law_1.01,0.365996789932251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,power_law_1.01,0.4713151931762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,power_law_1.01,0.5467520236968995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,power_law_1.01,0.7302656173706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,power_law_1.01,0.907692813873291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,power_law_1.01,1.2627008438110352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,power_law_1.01,1.645510482788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,power_law_1.2,0.6418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.4037631988525391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.45981440544128416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,balanced,0.029834667841593426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,balanced,0.03517866631348928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,balanced,0.03689600030581156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,balanced,0.037274666130542755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,balanced,0.05533866584300995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,balanced,0.09290132919947307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,balanced,0.12424533565839131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,balanced,0.15678399801254272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,balanced,0.20999999841054282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,balanced,0.2635466655095418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,balanced,0.03412266572316488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,balanced,0.03689600030581156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,balanced,0.04022933294375738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,balanced,0.07472533484299977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,balanced,0.08789333701133728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,power_law_1.2,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,power_law_1.2,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,power_law_1.2,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,power_law_1.2,0.05856639742851257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,power_law_1.2,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,power_law_1.2,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,power_law_1.2,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,power_law_1.2,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,power_law_1.2,0.11072640419006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,power_law_1.2,0.143449604511261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,power_law_1.2,0.1533184051513672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,power_law_1.2,0.21621758937835694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.11031039953231811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.1172287940979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.11885440349578857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.12465280294418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.1267840027809143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.12890880107879638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.13674880266189576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.1410688042640686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.14606720209121704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.15812480449676514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.17159039974212648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.19116159677505493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.21948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.2688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.3389695882797241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.41855359077453613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.5201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,power_law_1.2,0.7262015819549561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,power_law_1.2,0.260646390914917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,power_law_1.2,0.9102592468261719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.2288831710815429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,power_law_1.2,0.3797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,power_law_1.2,1.4601344108581542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,power_law_1.2,0.0991807997226715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,power_law_1.2,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,power_law_1.2,0.14543360471725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,power_law_1.2,0.18795520067214966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,power_law_1.2,0.20734078884124757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,power_law_1.2,0.27445120811462403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,power_law_1.2,0.28827519416809083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,power_law_1.2,0.28808319568634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,power_law_1.2,0.29144959449768065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,power_law_1.2,0.29159040451049806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,power_law_1.2,0.3054336071014404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,power_law_1.2,0.3076224088668823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,power_law_1.2,0.30561280250549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,power_law_1.2,0.3325632095336914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,power_law_1.2,0.338752007484436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,power_law_1.2,0.35863680839538575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,power_law_1.2,0.37500801086425783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,power_law_1.2,0.40508160591125486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,power_law_1.2,0.4450496196746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,power_law_1.2,0.5204480171203614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,power_law_1.2,0.5839615821838379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,power_law_1.2,0.7231296062469482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,power_law_1.2,0.7765247821807861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,power_law_1.2,1.0214271545410156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,power_law_1.2,1.3461376190185548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,power_law_1.2,1.6823616027832031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,power_law_1.2,0.5409855842590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,balanced,0.025653332471847534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,power_law_1.2,2.0943679809570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,balanced,0.025418666501839954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,balanced,0.027957332630952198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,balanced,0.029701332251230877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,balanced,0.029743999242782593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,balanced,0.030202666918436687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,balanced,0.0317493329445521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,balanced,0.02979733298222224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,balanced,0.03025600065787633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,balanced,0.03271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,balanced,0.03408000121514002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,balanced,0.0372533326347669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,balanced,0.03641066700220108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,balanced,0.037077332536379494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,balanced,0.039962666730086006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,balanced,0.060533334811528526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,balanced,0.07689066727956136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,balanced,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,balanced,0.10945066809654236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,power_law_1.2,0.6368512153625489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,power_law_1.2,0.9546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,balanced,0.07271466652552287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,balanced,0.10785067081451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,balanced,0.1323253313700358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,balanced,0.13334932923316956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,balanced,0.13310399651527405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,balanced,0.13457066814104715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,balanced,0.13486933708190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,balanced,0.13455466429392496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,balanced,0.13774399956067404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,balanced,0.13717333475748697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,balanced,0.13970133662223816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,balanced,0.14477333426475525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,balanced,0.14453333616256714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,balanced,0.15061333775520325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,balanced,0.16579733292261759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,balanced,0.1695573329925537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,balanced,0.22287466128667197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,balanced,0.22057066361109415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,balanced,0.298144002755483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,balanced,0.3205759922663371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,balanced,0.4559679826100667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,balanced,0.5308693250020345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,balanced,0.7795200347900391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,balanced,0.9378506342569987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.14986879825592042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.20232319831848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.3002432107925415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.32245121002197263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.33654398918151857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.34061439037323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.3506560087203979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.35779199600219724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.3687551975250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.3611520051956177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.3865983963012695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.3944256067276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.40244479179382325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.4204160213470459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.45711359977722166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.4650432109832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.01,0.5312640190124511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.01,0.6040063858032226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.01,0.7254079818725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.01,0.7489984035491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.01,0.9075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.01,1.0159104347229004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.01,1.3460927963256837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.01,1.6457792282104493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.0523904025554657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.09376639723777772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.11319680213928222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.12158080339431762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.1482751965522766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.18417919874191285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.24816639423370362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.3124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.41721601486206056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.5655871868133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,power_law_1.2,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,power_law_1.2,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,power_law_1.2,0.0427264004945755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,power_law_1.2,0.04338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,power_law_1.2,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,power_law_1.2,0.08047999739646912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,power_law_1.2,0.10330239534378052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,power_law_1.2,0.11296000480651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,power_law_1.2,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,power_law_1.2,0.18239359855651854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,power_law_1.2,0.24209280014038087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,power_law_1.2,0.3091327905654907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,power_law_1.2,0.4171135902404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,power_law_1.2,0.5739264011383056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.04953599870204926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.07727360129356384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.0877120018005371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.14594559669494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.17223039865493775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.22917120456695556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,balanced,0.05256533126036326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,balanced,0.06689066688219707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,balanced,0.09702400366465251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,balanced,0.14085333546002707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,balanced,0.17678932348887125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,balanced,0.18037333091100058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,balanced,0.18074132998784384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,balanced,0.1807360053062439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,balanced,0.18258132537206015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,balanced,0.1829599936803182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,balanced,0.1885546644528707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,balanced,0.18727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,balanced,0.19236266613006592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,balanced,0.200981338818868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,balanced,0.20145066579182944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,balanced,0.20946133136749268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,balanced,0.2304853399594625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,balanced,0.2495466669400533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,balanced,0.3596266508102417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,balanced,0.3304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,balanced,0.477946678797404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,balanced,0.5242933432261149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,balanced,0.763802687327067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,balanced,0.9041972955067953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,balanced,1.3364639282226562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,balanced,1.660421371459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.35133440494537355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,power_law_1.01,0.06096000075340271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,power_law_1.01,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,power_law_1.01,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,power_law_1.01,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,power_law_1.01,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,power_law_1.01,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,power_law_1.01,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,power_law_1.01,0.0892416000366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,power_law_1.01,0.09511680006980897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,power_law_1.01,0.09907199740409851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,power_law_1.01,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,power_law_1.01,0.12437119483947753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.01,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.01,0.17111680507659913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.01,0.1819391965866089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.01,0.226310396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.01,0.2969664096832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.01,0.39768960475921633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.01,0.5003647804260254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.01,0.7250688076019287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.01,0.9415488243103027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,power_law_1.2,0.43796482086181643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,power_law_1.2,0.5145664215087891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,balanced,0.04205333193143209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,balanced,0.05451733370621999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,balanced,0.07401066521803538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,balanced,0.07366933425267537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,balanced,0.07470400134722392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,balanced,0.07701333363850911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,balanced,0.07919466495513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,balanced,0.08392533659934998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,balanced,0.08741866548856099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,balanced,0.08994666735331218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,balanced,0.09744532903035481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,balanced,0.10777599612871806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,balanced,0.11931199828783672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,balanced,0.1418293317159017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,balanced,0.1616373360157013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,balanced,0.20100800196329752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,balanced,0.2524426579475403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,balanced,0.3340746561686198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,balanced,0.41838932037353516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,balanced,0.5949546496073405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,balanced,0.7579733530680338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.0423552006483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.06039680242538452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.11029759645462037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.1828287959098816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,power_law_1.01,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,power_law_1.01,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,power_law_1.01,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,power_law_1.01,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,power_law_1.01,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,power_law_1.01,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.01,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.01,0.16887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.07130879759788514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.01,0.2131903886795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.01,0.288703989982605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.01,0.4062655925750732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,power_law_1.01,0.09902080297470092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,power_law_1.01,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,power_law_1.01,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,power_law_1.01,0.11242879629135132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.01,0.6046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,power_law_1.01,0.1065343976020813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,power_law_1.01,0.10954879522323609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,power_law_1.01,0.11274240016937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,power_law_1.01,0.11830400228500366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,power_law_1.01,0.11759999990463257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,power_law_1.01,0.12037760019302368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,power_law_1.01,0.12199039459228515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,power_law_1.01,0.12716799974441528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,power_law_1.01,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,power_law_1.01,0.13242239952087403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,power_law_1.01,0.14947199821472168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,power_law_1.01,0.1784832000732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,power_law_1.01,0.1894271969795227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,power_law_1.01,0.2399679899215698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,power_law_1.01,0.23676159381866455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,power_law_1.01,0.3121216058731079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,power_law_1.01,0.34773759841918944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,power_law_1.01,0.4674111843109131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,power_law_1.01,0.6101823806762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,power_law_1.01,0.8134528160095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,power_law_1.01,1.109727954864502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.23280000686645508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.2439487934112549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.25065600872039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.01,0.8382464408874511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.01,1.1477439880371094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,power_law_1.01,0.4032127857208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.01,1.556985569000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,power_law_1.01,0.4532800197601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.2,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.2,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.2,0.12314239740371705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.2,0.1497215986251831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.2,0.2148672103881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.2,0.22053759098052977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.2,0.2303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.2,0.2327039957046509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.2,0.2397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.2,0.24328320026397704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.2,0.24975359439849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.2,0.25932159423828127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.2,0.2716223955154419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.2,0.28577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.2,0.2974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.2,0.31413118839263915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.2,0.3550976037979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.2,0.4012735843658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.2,0.4974976062774658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.2,0.595366382598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.2,0.7761087894439698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.2,0.9354816436767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.2,1.2702719688415527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,balanced,0.11768000324567159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,balanced,0.1856586734453837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,balanced,0.32253867387771606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,balanced,0.5946026643117269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,balanced,1.1182186603546143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.2,1.5196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,balanced,1.6264479955037434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,balanced,1.634149392445882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,balanced,1.6422239939371746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.2,2.1755840301513674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,balanced,1.6444212595621746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,balanced,1.6508266131083171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,balanced,1.6604480743408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,balanced,1.675381342569987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,balanced,1.6785119374593098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,balanced,1.6995946566263835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,balanced,1.7184373537699382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,balanced,1.7330880165100098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,balanced,1.7571733792622883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,balanced,2.078810691833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,balanced,1.870757261912028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.2,2.7225343704223635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,balanced,2.361839930216471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,balanced,2.1075679461161294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,balanced,3.0482559204101562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,balanced,2.5719359715779624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,balanced,3.970032056172689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,balanced,3.889418601989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,balanced,6.000538508097331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,balanced,6.4155839284261065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,power_law_1.01,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,power_law_1.01,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,power_law_1.01,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,power_law_1.01,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,power_law_1.2,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,power_law_1.2,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,power_law_1.2,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,power_law_1.2,0.022867199778556824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,power_law_1.2,0.023647999763488768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,power_law_1.2,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,power_law_1.2,0.026521599292755126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,power_law_1.2,0.029094401001930236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,power_law_1.2,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.01,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.01,0.07996799945831298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.15000319480895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.01,0.10176000595092774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.01,0.14632320404052734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.16554239988327027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.01,0.1696768045425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.27693440914154055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.01,0.23761279582977296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.01,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.01,0.12019840478897095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.2989311933517456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.01,0.12007039785385132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.01,0.1524608016014099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.01,0.1769152045249939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.01,0.24372479915618897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.01,0.22337279319763184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.01,0.22468481063842774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.01,0.22583041191101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.01,0.23375999927520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.01,0.24618239402770997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.01,0.24673280715942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.01,0.24472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.01,0.25715200901031493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.01,0.2718656063079834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.01,0.28901119232177735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.01,0.2825344085693359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.01,0.31266560554504397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.01,0.3423487901687622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.01,0.40250239372253416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.01,0.45937280654907225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.01,0.5878655910491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.01,0.6465407848358155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,power_law_1.2,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.01,0.8494015693664551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.01,1.0276927947998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,power_law_1.2,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,power_law_1.2,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.01,1.4322239875793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.01,1.7604799270629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,power_law_1.2,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,power_law_1.2,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,power_law_1.2,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,power_law_1.2,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,power_law_1.2,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,power_law_1.2,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,power_law_1.2,0.08557440042495727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,power_law_1.2,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,power_law_1.2,0.11093120574951172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,power_law_1.2,0.1360576033592224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,power_law_1.2,0.15930240154266356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,power_law_1.2,0.21766400337219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,power_law_1.2,0.2632960081100464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,power_law_1.2,0.355513596534729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,power_law_1.2,0.46788477897644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,power_law_1.2,0.7018176078796386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,power_law_1.2,0.9044095993041992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.01,0.31217920780181885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,balanced,0.037258667250474296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,balanced,0.07336000104745229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,balanced,0.07458666463692983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,balanced,0.0745600014925003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,balanced,0.07558399935563405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,balanced,0.0757173349459966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,balanced,0.0775679995616277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,balanced,0.07813866436481476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,balanced,0.07889066636562347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,balanced,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,balanced,0.08666666348775227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,balanced,0.0928106705347697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,balanced,0.09912000099817912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.10578133662541707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.12149866422017415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.13474667072296143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.1731520096460978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,power_law_1.2,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.19774933656056723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.2743786573410034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.33949331442515057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.4876319964726766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,power_law_1.2,0.06046720147132874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.6298453410466512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,power_law_1.2,0.05982720255851746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,power_law_1.2,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,power_law_1.2,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,power_law_1.2,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,power_law_1.2,0.07900800108909607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,power_law_1.2,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,power_law_1.2,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,power_law_1.2,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,power_law_1.2,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,power_law_1.2,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,power_law_1.2,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.2,0.13842560052871705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.2,0.1575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.2,0.19756799936294556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.2,0.25080959796905516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.2,0.33845760822296145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.2,0.4564544200897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.2,0.6924032211303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.2,0.8845631599426269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.04082559943199158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.04433279931545257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.10333440303802491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.16819839477539061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.23816959857940673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.06932479739189149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.08073599934577942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.1014143943786621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.10094079971313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.10371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.10781439542770385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.11288319826126099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.11930240392684936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.128057599067688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.13419519662857055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.14264960289001466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.15582079887390138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.17466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.2067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.2502336025238037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.3162175893783569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.40471038818359373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.4986623764038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,power_law_1.2,0.6491199970245362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,power_law_1.2,0.9179264068603515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,power_law_1.2,1.1393919944763184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,power_law_1.01,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,power_law_1.01,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,power_law_1.01,0.11857919692993164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,power_law_1.01,0.1219264030456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,power_law_1.01,0.11018879413604736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,power_law_1.01,0.1283136010169983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,power_law_1.01,0.1183359980583191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,power_law_1.01,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,power_law_1.01,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,power_law_1.01,0.12862080335617065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,power_law_1.01,0.12751359939575196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,power_law_1.01,0.1298367977142334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,power_law_1.01,0.12884479761123657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,power_law_1.01,0.14238719940185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,power_law_1.01,0.14194560050964355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,power_law_1.01,0.14470399618148805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,power_law_1.01,0.1615679979324341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.01,0.17383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.01,0.20524799823760986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.01,0.23223679065704345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.01,0.29427199363708495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.01,0.3861567974090576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.01,0.5097856044769287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.01,0.6506303787231446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.01,1.0116543769836426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.01,1.3459199905395507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.1949056029319763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.10704640150070191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.21178879737854003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.13409279584884642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.13711999654769896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,power_law_1.01,0.34314239025115967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.1942720055580139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.25674240589141845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.295199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,power_law_1.01,0.43016958236694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,power_law_1.2,0.3993407964706421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,power_law_1.2,0.5142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,power_law_1.01,0.6122496128082275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,power_law_1.2,0.7100992202758789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,power_law_1.01,0.8885760307312012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,balanced,0.14386666814486185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,balanced,0.239519993464152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,balanced,0.43273599942525226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,balanced,0.8231626351674398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,balanced,1.5966506004333496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,balanced,3.141098658243815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,balanced,3.1466506322224936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,balanced,3.1528425216674805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,balanced,3.1607840855916343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,balanced,3.163914680480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,balanced,3.176080067952474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,balanced,3.1828746795654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,balanced,3.195669492085775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,balanced,3.2157920201619468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,balanced,3.2371412913004556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,balanced,3.2468105951944985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,balanced,3.2831252415974936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,balanced,3.332928021748861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,balanced,3.3961493174235025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,balanced,3.524112065633138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,balanced,3.617472012837728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,power_law_1.2,0.9104703903198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,power_law_1.2,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,balanced,3.7709385553995767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,power_law_1.2,0.05692800283432007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,power_law_1.2,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,balanced,4.077237447102864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,power_law_1.2,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,balanced,4.553999900817871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,power_law_1.2,0.11096960306167603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,balanced,4.888949394226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,power_law_1.2,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,power_law_1.2,0.14293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,balanced,6.6268157958984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,power_law_1.2,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,power_law_1.2,0.15479040145874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,power_law_1.2,0.15866880416870116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,balanced,7.26911989847819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,power_law_1.2,0.16626559495925902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,power_law_1.2,0.17114880084991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,power_law_1.2,0.1931264042854309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,power_law_1.2,0.20426878929138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,power_law_1.2,0.20830080509185792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,power_law_1.2,0.23374719619750978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,power_law_1.2,0.26449921131134035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,power_law_1.2,0.3282880067825317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,power_law_1.2,0.3779583930969238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,power_law_1.2,0.49245438575744627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,power_law_1.2,0.6170688152313233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,power_law_1.2,0.8574336051940918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,power_law_1.2,1.0434816360473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,power_law_1.2,1.457260799407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,power_law_1.2,1.8529727935791016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,balanced,0.0625600020090739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,balanced,0.06409599880377452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,balanced,0.0747573326031367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,balanced,0.09247466921806335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,balanced,0.11986133456230164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,balanced,0.16482667128245035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,balanced,0.16172800461451212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,balanced,0.16226133704185486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,balanced,0.16459199786186218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,balanced,0.1665066679318746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,balanced,0.16548267006874084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,balanced,0.16617600123087564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,balanced,0.16311466693878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,balanced,0.16473066806793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,balanced,0.1662773291269938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,balanced,0.1715466578801473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,balanced,0.17492266496022543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,balanced,0.18430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,balanced,0.19563200076421103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,balanced,0.2108853260676066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,balanced,0.2372373342514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,balanced,0.2606559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,balanced,0.3148159980773926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,balanced,0.4041279951731364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,balanced,0.4949013392130534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,balanced,0.6358240048090616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.24822399616241456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.4334400177001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.8243519783020019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.2964735984802247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.8556991577148438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.2325183868408205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.0596160888671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.220403289794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.334956741333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.440403366088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.581612777709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,power_law_1.2,3.6392383575439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,power_law_1.2,3.7838783264160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,power_law_1.2,3.886598587036133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,power_law_1.2,3.9108608245849608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,power_law_1.2,4.08271369934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,power_law_1.2,4.375852966308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,power_law_1.2,4.798112106323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,power_law_1.2,5.387372970581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,power_law_1.2,5.668870544433593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,power_law_1.2,6.527193450927735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,power_law_1.2,6.139475250244141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,power_law_1.2,7.259954833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,power_law_1.2,8.257234954833985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,power_law_1.2,9.920819091796876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.01,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.01,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.01,0.0559935986995697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,power_law_1.2,12.235372924804688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.01,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.01,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.01,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.01,0.061740797758102414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.01,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.01,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.01,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.01,0.1255231976509094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.01,0.15084799528121948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.01,0.19464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.01,0.20468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.01,0.25841920375823973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.01,0.31048319339752195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.01,0.4273856163024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.01,0.5185599803924561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,balanced,0.033957332372665405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,balanced,0.03698666642109553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,balanced,0.08271466692288716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,balanced,0.11120532949765523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,balanced,0.12994666894276938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,balanced,0.18466667334238687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,balanced,0.22635199626286825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,balanced,0.3229440053304036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,balanced,0.4368160168329875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.2,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.2,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.2,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.2,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.2,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.2,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.2,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.2,0.12637439966201783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.2,0.16625280380249025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.2,0.20892159938812255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.2,0.32010879516601565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.2,0.32448639869689944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,power_law_1.2,0.07711359858512878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,power_law_1.2,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,power_law_1.2,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,power_law_1.2,0.11790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.15196160078048707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.21028480529785157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,power_law_1.2,0.09628159999847412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.25788800716400145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,power_law_1.2,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3414655923843384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,power_law_1.2,0.12028800249099732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.3583807945251465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,power_law_1.2,0.12337919473648071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.3735872030258179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,power_law_1.2,0.11925760507583619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.37934079170227053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,power_law_1.2,0.12232320308685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.39758079051971434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,power_law_1.2,0.12981760501861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.41182718276977537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.4197055816650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.43248639106750486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.4492479801177979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.4604800224304199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.49272961616516114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,power_law_1.2,0.13165440559387206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5484992027282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,power_law_1.2,0.13228800296783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.574944019317627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,power_law_1.2,0.14153599739074707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.6830336093902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,power_law_1.2,0.14670720100402831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.7773439884185791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,power_law_1.2,0.14935040473937988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,power_law_1.01,0.9634559631347657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,power_law_1.2,0.1662976026535034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.1096320152282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.2,0.17479679584503174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.2,0.2216576099395752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.437235164642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.2,0.24490880966186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.7435903549194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.2,0.35250558853149416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.2,0.4480127811431885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.3929792404174806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.2,0.5932544231414795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.2,0.7350719928741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.0230783462524413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.2,1.1160832405090333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.2,1.5432127952575683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,2,power_law_1.01,0.1593791961669922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,2,power_law_1.01,0.2081216096878052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,2,power_law_1.01,0.24377601146697997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,2,power_law_1.01,0.3369152069091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,2,power_law_1.01,0.3765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,2,power_law_1.01,0.573798418045044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,2,power_law_1.01,0.6016640186309814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,2,power_law_1.01,0.6067647933959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,2,power_law_1.01,0.6346303939819335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,2,power_law_1.01,0.6369088172912598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,2,power_law_1.01,0.6406400203704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,2,power_law_1.01,0.6520512104034424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,2,power_law_1.01,0.6886079788208008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,2,power_law_1.01,0.6659200191497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,2,power_law_1.01,0.6878528118133544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,2,power_law_1.01,0.7092607975006103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,2,power_law_1.01,0.7181695938110352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,2,power_law_1.01,0.7703423976898194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.01,0.8019007682800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.01,0.9087424278259277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.01,1.0067392349243165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.01,1.2155776023864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.01,1.5060864448547364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.01,1.9737472534179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.01,2.5261375427246096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.01,3.617036819458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.01,4.519583892822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,power_law_1.01,0.10307199954986572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,power_law_1.01,0.09549440145492553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,power_law_1.01,0.07857919931411743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,power_law_1.01,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,power_law_1.01,0.0958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,power_law_1.01,0.09278079867362976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,power_law_1.01,0.09502720236778259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,power_law_1.01,0.10337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,power_law_1.01,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,power_law_1.01,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,power_law_1.01,0.11364480257034301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,power_law_1.01,0.11942399740219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,power_law_1.01,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,power_law_1.01,0.13662079572677613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,power_law_1.01,0.14440959692001343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,power_law_1.01,0.17242239713668822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,power_law_1.01,0.20776960849761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,power_law_1.01,0.2503743886947632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,power_law_1.01,0.30617599487304686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,power_law_1.01,0.4144768238067627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,power_law_1.01,0.5239744186401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,power_law_1.01,0.7550015926361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,power_law_1.01,0.9791232109069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,balanced,0.04459733267625173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,balanced,0.07062933345635732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,balanced,0.09676800171534221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,balanced,0.14731733004252115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,balanced,0.19710934162139893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,balanced,0.19977066914240518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,balanced,0.20030933618545532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,balanced,0.20018132527669272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,balanced,0.20114666223526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,balanced,0.20165866613388062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,balanced,0.20324265956878662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,balanced,0.20482667287190756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,balanced,0.20860799153645834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,balanced,0.21061333020528158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,balanced,0.21474132935206094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,balanced,0.2233546574910482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,balanced,0.2298346757888794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,balanced,0.24514132738113403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,balanced,0.2605280081431071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,balanced,0.29069334268569946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,balanced,0.3210560083389282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,balanced,0.37486398220062256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,balanced,0.4509066740671794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,balanced,0.5554986794789633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,balanced,0.7063360214233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,power_law_1.01,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,power_law_1.01,0.02802560031414032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,power_law_1.01,0.026495999097824095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,power_law_1.01,0.029407998919486998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,power_law_1.01,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,power_law_1.01,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,power_law_1.01,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,power_law_1.01,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,power_law_1.01,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,power_law_1.01,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,power_law_1.01,0.11559040546417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,power_law_1.01,0.13493759632110597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,power_law_1.01,0.20057599544525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,power_law_1.01,0.2576319932937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.06847360134124755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.1307904005050659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.14732799530029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.27112319469451907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.343507194519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.10683519840240478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.4573184013366699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.11914240121841431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,balanced,0.04204800228277842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,balanced,0.04299733539422353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,balanced,0.07362666726112366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,balanced,0.09895466764767964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,balanced,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,balanced,0.1014400025208791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,balanced,0.10314133763313293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,balanced,0.10317333539326985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,balanced,0.10415466626485188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,balanced,0.10700266559918721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,balanced,0.10545066992441814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,balanced,0.10909866293271382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,balanced,0.11450133721033733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,balanced,0.11556800206502278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.1468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,balanced,0.12106133500734965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,balanced,0.13237866759300232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,balanced,0.13804800311724344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,balanced,0.15706666310628256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,balanced,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,balanced,0.219925324122111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,balanced,0.253109335899353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,balanced,0.3390186627705892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,balanced,0.4096213181813558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,balanced,0.5705013275146484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,balanced,0.7167893250783285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.15960960388183593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.22495999336242675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.6457920074462891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.30782721042633054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.46329598426818847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,0.885478401184082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.5310143947601318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,power_law_1.2,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,power_law_1.2,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,power_law_1.2,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,power_law_1.2,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,power_law_1.2,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,power_law_1.2,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,power_law_1.2,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,power_law_1.2,0.1177024006843567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,power_law_1.2,0.15152000188827514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,power_law_1.2,0.17689599990844726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,power_law_1.2,0.24148480892181395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,power_law_1.2,0.32446720600128176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,power_law_1.2,0.4585984230041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,power_law_1.01,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,power_law_1.01,0.11423360109329224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,power_law_1.01,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,power_law_1.01,0.12796159982681274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,power_law_1.01,0.134278404712677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,power_law_1.01,0.1315775990486145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,power_law_1.01,0.136953604221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,power_law_1.01,0.13957120180130006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,power_law_1.01,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,power_law_1.01,0.16376960277557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,power_law_1.01,0.17358720302581787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.01,0.1887488007545471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.01,0.2205888032913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.01,0.24557440280914306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.01,0.30420479774475095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.01,0.3494080066680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.01,0.4763967990875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.01,0.6100736141204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.01,0.8194560050964356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.01,1.0473535537719727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.01,0.11695359945297241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.01,0.1464192032814026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.01,0.14327679872512816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.01,0.1534208059310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.01,0.15286400318145751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.01,0.1429759979248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.01,0.13895679712295533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.01,0.1447424054145813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.01,0.1456768035888672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.01,0.14335360527038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.01,0.1541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.01,0.15055999755859376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.01,0.1534656047821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.01,0.15707520246505738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.01,0.1676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.01,0.16879359483718873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.01,0.18099199533462523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.01,0.20396800041198732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.01,0.21666560173034669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.01,0.2617408037185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,power_law_1.2,0.5845759868621826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.01,0.30770559310913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.01,0.38915839195251467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.01,0.3943808078765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.01,0.48151679039001466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.01,0.614847993850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.01,0.8723135948181152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.01,0.9931455612182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.12171519994735717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.16402560472488403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.21702399253845214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.270796799659729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.3110719919204712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.403001594543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.4181056022644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.435968017578125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.44508800506591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.463699197769165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.4707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.4920639991760254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.5052288055419922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.523801612854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.5366528034210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6034304141998291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.6473983764648438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,0.7532415866851807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,0.8528960227966309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,1.0547712326049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,1.221664047241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,1.606175994873047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.2,1.999001693725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.2,2.7566848754882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.2,3.636697769165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.09088640213012696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.13400319814682007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.13597439527511596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.1363263964653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.18176640272140504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.20974080562591552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.27930240631103515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.36431360244750977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,power_law_1.2,0.5138495922088623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,power_law_1.2,0.6416768074035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,power_law_1.01,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,power_law_1.01,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,power_law_1.01,0.08657919764518737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,power_law_1.01,0.10165120363235473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,power_law_1.01,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,power_law_1.01,0.1112064003944397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,power_law_1.01,0.1129472017288208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,power_law_1.01,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,power_law_1.01,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,power_law_1.01,0.1356287956237793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,power_law_1.01,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,power_law_1.01,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,power_law_1.01,0.15607680082321168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,power_law_1.01,0.17965439558029175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,power_law_1.01,0.20602879524230958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,power_law_1.01,0.2561471939086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,power_law_1.01,0.3087935924530029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,power_law_1.01,0.42169599533081054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,power_law_1.01,0.5198976039886475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,power_law_1.01,0.7287871837615967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,power_law_1.01,0.9532352447509765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,power_law_1.01,1.2868736267089844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,power_law_1.01,1.7563583374023437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,power_law_1.2,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,power_law_1.2,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,power_law_1.2,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,power_law_1.2,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,power_law_1.2,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,power_law_1.2,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,power_law_1.2,0.04993920028209686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,power_law_1.2,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,power_law_1.2,0.13462400436401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,power_law_1.2,0.1619647979736328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,power_law_1.2,0.25427200794219973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,power_law_1.2,0.39548799991607664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,power_law_1.2,0.5098624229431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,power_law_1.2,0.751692819595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.14941439628601075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.17519999742507936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.16737279891967774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.16195839643478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.138374400138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.16631679534912108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.16549760103225708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.17818880081176758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.16844799518585205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.17944960594177245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.1792448043823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.18032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.1812351942062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.18554879426956178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.19306880235671997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.19447040557861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.21473278999328613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.2553600072860718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.24645121097564698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.29320321083068845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.347705602645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,power_law_1.01,0.4167168140411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,power_law_1.01,0.43657598495483396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,power_law_1.01,0.6080895900726319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,power_law_1.01,0.7496575832366943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,power_law_1.01,1.0139519691467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,power_law_1.01,1.3384703636169433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.2,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.2,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.2,0.13099520206451415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.2,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.2,0.17076480388641357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.2,0.1813696026802063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.2,0.16590720415115356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.2,0.17029759883880616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.2,0.1746559977531433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.2,0.1870784044265747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.2,0.18751360177993776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.2,0.19077119827270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.2,0.1952831983566284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.2,0.2046976089477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.2,0.20264959335327148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.2,0.21338241100311278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.2,0.22901759147644044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.2,0.2566911935806274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.2,0.28908159732818606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.2,0.3573760032653809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.2,0.39976959228515624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.2,0.5213247776031494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.2,0.591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.2,0.7910399913787842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.2,0.9960960388183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.2,1.3282815933227539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.2,1.870591926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,balanced,0.05258133510748545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,balanced,0.05625600119431814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,balanced,0.13868799805641174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,balanced,0.15781333049138388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,balanced,0.1611253321170807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,balanced,0.16357866923014322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,balanced,0.16517866651217142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,balanced,0.16698666413625082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,balanced,0.16926399866739908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,balanced,0.17125866810480753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,balanced,0.17320533593495688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,balanced,0.17806933323542276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,balanced,0.18405866622924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,balanced,0.18722132841746011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,balanced,0.1969333291053772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,balanced,0.2197493314743042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.23813333113988241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.2833813428878784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.3131840030352275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.4323413372039795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.5007733503977457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.7017066478729248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.8888053099314371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,balanced,1.304800033569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.637386639912923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,balanced,0.04012800008058548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,balanced,0.03926933308442434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,balanced,0.03998400022586187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,balanced,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,balanced,0.05425600210825602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,balanced,0.06126933296521505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,balanced,0.07175466914971669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.08054933448632558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.10325866937637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.12059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.1530400017897288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.17776000499725342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.25082133213679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.3201013406117757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.46089601516723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.5978293418884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,power_law_1.2,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,power_law_1.2,0.15928319692611695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,power_law_1.2,0.21749119758605956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,power_law_1.2,0.2804608106613159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,power_law_1.2,0.4110527992248535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,power_law_1.2,0.42091522216796873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,power_law_1.2,0.43427839279174807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,power_law_1.2,0.4514048099517822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,power_law_1.2,0.46752638816833497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,power_law_1.2,0.4946752071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,power_law_1.2,0.4812607765197754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,power_law_1.2,0.5135488033294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,power_law_1.2,0.5353151798248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,power_law_1.2,0.5357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,power_law_1.2,0.5740287780761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,power_law_1.2,0.626527976989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,power_law_1.2,0.6314752101898193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,power_law_1.2,0.727129602432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,power_law_1.2,0.7432703971862793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,power_law_1.2,0.8650943756103515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,power_law_1.2,0.9705727577209473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,power_law_1.2,1.2103232383728026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,power_law_1.2,1.4583295822143554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,power_law_1.2,1.923891258239746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,power_law_1.2,2.4206783294677736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,balanced,0.02972800036271413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,balanced,0.03922666609287262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,balanced,0.04437866806983948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,balanced,0.08694932858149211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,balanced,0.10965866843859355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,balanced,0.12494933605194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,balanced,0.15727999806404114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,balanced,0.20117332537968954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,balanced,0.2721280058224996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,balanced,0.3362720012664795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.01,0.1205183982849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.01,0.19120639562606812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.01,0.2511104106903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.01,0.38265600204467776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.01,0.38071041107177733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.01,0.3940927982330322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.01,0.39685759544372556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.01,0.4067903995513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.01,0.4068352222442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.01,0.4313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.01,0.43829760551452634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.01,0.43909120559692383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.01,0.46569600105285647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.01,0.47302398681640623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.01,0.5070911884307862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.01,0.5813375949859619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.01,0.5903168201446534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.01,0.7001152038574219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.01,0.7764607906341553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.01,0.9739456176757812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.01,1.052889633178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.01,1.3217087745666505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.01,1.7597312927246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.01,2.459903907775879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.01,3.045734405517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,balanced,0.04418666660785675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,balanced,0.06718933085600536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,balanced,0.09591466188430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,balanced,0.14918399850527445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,balanced,0.24673600991566977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,balanced,0.347925345102946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,balanced,0.3433386484781901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,balanced,0.3466399908065796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,balanced,0.3463946580886841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,balanced,0.349455992380778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,balanced,0.3513706525166829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,balanced,0.350874662399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,balanced,0.3558559815088908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,balanced,0.36475733915964764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,balanced,0.36882134278615314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,balanced,0.38042132059733075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,balanced,0.39025068283081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,balanced,0.40462398529052734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,balanced,0.4342506726582845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,balanced,0.4583306709925334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,balanced,0.5107040007909139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,balanced,0.5669493277867635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,balanced,0.6680586338043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,balanced,0.803658644358317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,balanced,1.0009386539459229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,balanced,1.2942186991373699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.12437759637832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,power_law_1.01,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.1404096007347107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,power_law_1.01,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,power_law_1.01,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,power_law_1.01,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,power_law_1.01,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,power_law_1.01,0.07333760261535645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.01,0.08643839955329895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.01,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.01,0.12003200054168701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.01,0.14921599626541138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.01,0.1803455948829651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.01,0.22961280345916749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.01,0.2784960031509399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.01,0.39752318859100344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.01,0.510099220275879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.20044798851013185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.03932799994945526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.04993920028209686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.12353279590606689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.16549760103225708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.2458751916885376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.22450559139251708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.24769918918609618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.05994240045547485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.09304959774017334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.09329280257225037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.1300480008125305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.15525120496749878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.3680639982223511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.20936961174011232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.3827264070510864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.21793920993804933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,power_law_1.01,0.33752319812774656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,power_law_1.01,0.3993535995483398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,power_law_1.01,0.575110387802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,power_law_1.01,0.756883192062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.493177604675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.4519680023193359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.06370559930801392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.12333439588546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.14696320295333862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.17946239709854125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.20874240398406982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.30511999130249023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.3625407934188843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.4865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.6247935771942139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,power_law_1.01,1.1583104133605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,power_law_1.01,1.5217151641845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,2,power_law_1.2,0.1264448046684265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,2,power_law_1.2,0.13220479488372802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,2,power_law_1.2,0.1625983953475952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,2,power_law_1.2,0.18918399810791015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,2,power_law_1.2,0.23768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,2,power_law_1.2,0.258188796043396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,2,power_law_1.2,0.28828160762786864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,2,power_law_1.2,0.3576319932937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,2,power_law_1.2,0.3723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,2,power_law_1.2,0.3753727912902832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,2,power_law_1.2,0.3988607883453369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,2,power_law_1.2,0.4322624206542969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,2,power_law_1.2,0.4178624153137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,2,power_law_1.2,0.4109951972961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,2,power_law_1.2,0.4486080169677734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,2,power_law_1.2,0.4620096206665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,power_law_1.2,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,2,power_law_1.2,0.47617278099060056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,power_law_1.2,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,2,power_law_1.2,0.5144383907318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,2,power_law_1.2,0.5718400001525878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,power_law_1.2,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,power_law_1.2,0.09852160215377807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,2,power_law_1.2,0.6647168159484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,2,power_law_1.2,0.7417664051055908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,power_law_1.2,0.11749759912490845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,2,power_law_1.2,0.8927871704101562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,power_law_1.2,0.12078720331192017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,power_law_1.2,0.12409600019454955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,2,power_law_1.2,1.0667008399963378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,power_law_1.2,0.12458239793777466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,power_law_1.2,0.12979840040206908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,power_law_1.2,0.13342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,2,power_law_1.2,1.4156224250793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,power_law_1.2,0.14046720266342164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,balanced,0.04529066880544027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,balanced,0.04542933404445648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,power_law_1.2,0.14513280391693115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,balanced,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,balanced,0.08191466828187306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,power_law_1.2,0.15425920486450195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,2,power_law_1.2,1.7136127471923828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,balanced,0.08526933193206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,power_law_1.2,0.17428480386734008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.2,0.19132800102233888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.2,0.23175039291381835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,2,power_law_1.2,2.4570751190185547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.2,0.27123839855194093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.2,0.35015039443969725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,balanced,0.08471999565760295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.2,0.4353856086730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,balanced,0.0869653324286143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,balanced,0.08726933598518372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,balanced,0.09122666716575623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,2,power_law_1.2,3.0745471954345702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,balanced,0.09476799766222636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,balanced,0.09858666857083638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.2,0.5841472148895264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,balanced,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,balanced,0.11369066437085469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,balanced,0.11225600043932597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,balanced,0.13640532890955606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.2,0.7586048126220704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,balanced,0.1422826647758484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,balanced,0.18381333351135254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,balanced,0.20787199338277182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,balanced,0.2713279922803243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.2,1.0498047828674317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,balanced,0.3136426607767741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,balanced,0.46349867184956867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,balanced,0.5580586592356364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.2,1.472710418701172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,balanced,0.05044800043106079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,balanced,0.05755733450253805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,balanced,0.11050132910410564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,balanced,0.1704053282737732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,balanced,0.22741333643595377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,balanced,0.2353973388671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,balanced,0.23610132932662964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,balanced,0.23557867606480917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,balanced,0.23779733975728354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,balanced,0.23873066902160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,balanced,0.24329600731531778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,balanced,0.2511253356933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,balanced,0.25616000096003216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,balanced,0.26944533983866376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,balanced,0.29629333813985187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,balanced,0.310261329015096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,balanced,0.3563786745071411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,balanced,0.4080586830774943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,balanced,0.48438934485117596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,balanced,0.5816213289896647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,balanced,0.7479466597239176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,balanced,0.9681333700815836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,balanced,1.3033119837443035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,balanced,1.7070186932881672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,power_law_1.2,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,balanced,0.03618666778008143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,balanced,0.05260799825191498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,balanced,0.05434666574001312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,power_law_1.2,0.10078719854354859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,balanced,0.05635733405749003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,balanced,0.06633600095907848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,balanced,0.06623466809590657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,balanced,0.070783997575442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,balanced,0.07664533456166585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,balanced,0.0990559955437978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,balanced,0.11425066987673442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,balanced,0.15372266372044882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,balanced,0.1792693336804708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,balanced,0.24340800444285074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,balanced,0.31148799260457355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,balanced,0.44724265734354657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,balanced,0.5939626693725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,power_law_1.2,0.12281600236892701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,power_law_1.2,0.1583616018295288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,power_law_1.2,0.16444159746170045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,power_law_1.2,0.1681663990020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,power_law_1.2,0.1743872046470642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,power_law_1.2,0.18346879482269288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,power_law_1.2,0.19120639562606812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,power_law_1.2,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,power_law_1.2,0.20493440628051757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,power_law_1.2,0.21464319229125978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,power_law_1.2,0.22832000255584717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,power_law_1.2,0.24067199230194092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,power_law_1.2,0.2784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,power_law_1.2,0.30210559368133544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,power_law_1.2,0.36931838989257815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,power_law_1.2,0.41845121383666994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,power_law_1.2,0.5403135776519775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,power_law_1.2,0.6434751987457276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.07023360133171082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,power_law_1.2,0.871731185913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.09365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.0980351984500885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,power_law_1.2,1.0934335708618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.13602559566497802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.16510720252990724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,power_law_1.2,1.5297792434692383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.24620800018310546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.2749631881713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,power_law_1.2,1.965273666381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,balanced,0.052469333012898765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,balanced,0.05550933380921682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,balanced,0.061018665631612144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,balanced,0.07974400122960408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,balanced,0.12408000230789185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,balanced,0.17818133036295572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,balanced,0.1822026570638021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,balanced,0.18303465843200684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,balanced,0.18589866161346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,balanced,0.18767466147740683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,balanced,0.19263466199239096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,balanced,0.19829867283503214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,balanced,0.19986667235692343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,balanced,0.20794665813446045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,balanced,0.21541333198547363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,balanced,0.22418133417765299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,balanced,0.2437973419825236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,balanced,0.2812266747156779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,balanced,0.31138134002685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,balanced,0.3848693370819092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,balanced,0.44365866978963214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,balanced,0.6599680185317993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,balanced,0.7554506460825602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,balanced,1.1264586448669434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,balanced,1.4057920773824055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,balanced,2.076277256011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,balanced,2.743776003519694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,balanced,0.025600001215934753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,balanced,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,balanced,0.02536533276240031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,balanced,0.02980799973011017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,balanced,0.03292799989382426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,balanced,0.03307733436425527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,balanced,0.03346666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,balanced,0.03480533262093862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,balanced,0.07746133208274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,balanced,0.1055519978205363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,balanced,0.12748266259829202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,balanced,0.1673120061556498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,balanced,0.19835732380549112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,balanced,0.03176533430814743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,balanced,0.03256533294916153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,balanced,0.05705599983533224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,balanced,0.05709866682688395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,balanced,0.06011733412742615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,balanced,0.05801066756248474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,balanced,0.059808000922203064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,balanced,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,balanced,0.08719999591509502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,balanced,0.08650133013725281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,balanced,0.11031466722488403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,balanced,0.12418133020401001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,balanced,0.1662613352139791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,balanced,0.18820265928904215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,balanced,0.24626133839289346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,balanced,0.30434133609135944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.01,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.01,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.01,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.01,0.07641599774360656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.01,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.01,0.07816320061683654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.01,0.07728639841079712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.01,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.01,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.01,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.01,0.10221439599990845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.01,0.11787519454956055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.01,0.1244096040725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.01,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.01,0.1843392014503479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.01,0.24195199012756347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.01,0.2538815975189209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.01,0.3258879899978638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.01,0.4094079971313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.01,0.5661119937896728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.01,0.6800384044647216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.2,0.1207808017730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.2,0.1451840043067932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.2,0.14935679435729982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.2,0.13791359663009645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.2,0.13912960290908813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.2,0.1332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.2,0.13880959749221802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.2,0.14622080326080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.2,0.14973440170288085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.2,0.1505728006362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.2,0.15311360359191895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.2,0.1573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.2,0.16755199432373047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.2,0.17126400470733644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.2,0.17829760313034057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.2,0.20456318855285643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.2,0.22218239307403564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.2,0.28089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.2,0.30721919536590575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.2,0.3966399908065796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.2,0.38798720836639405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.2,0.518662405014038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.2,0.6705023765563964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.2,1.0218111991882324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.2,1.0735744476318358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,power_law_1.2,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,power_law_1.2,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,power_law_1.2,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,power_law_1.2,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,power_law_1.2,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,power_law_1.2,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,power_law_1.2,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.2,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.2,0.09260799884796142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.2,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.2,0.15192960500717162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.2,0.20538880825042724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.2,0.27834880352020264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.2,0.39801599979400637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.2,0.49628801345825196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.03968639969825745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.1158911943435669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.16110719442367555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.2076416015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.23427200317382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.315500807762146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,power_law_1.01,0.4526400089263916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.09645439982414246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.08874239921569824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.09503359794616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.10029439926147461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.10343680381774903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.10833920240402221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.11607680320739747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.13516160249710082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.1549056053161621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.18131840229034424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.22305281162261964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.26044158935546874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.33511040210723875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.40988798141479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.6011839866638183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,0.7556416034698487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.01,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.01,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.01,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.01,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.01,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.01,0.08586879968643188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.01,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.01,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.01,0.10357760190963745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.01,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.01,0.12339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.01,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.01,0.16272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.01,0.21580159664154053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.01,0.24770560264587402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.01,0.3146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.01,0.376531195640564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.01,0.5058688163757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.01,0.5718016147613525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,balanced,0.05165866514046987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,balanced,0.055120001236597695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,balanced,0.06819733480612437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,balanced,0.0805920014778773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,balanced,0.082096000512441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,balanced,0.08101333181063335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,balanced,0.08060800035794576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,balanced,0.08468799789746602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,balanced,0.08367466926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08503466844558716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,balanced,0.09291733304659526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,balanced,0.0939626693725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,balanced,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,balanced,0.10529067118962605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.1143893301486969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.14566933115323386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.14695466558138529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.18545599778493246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.21383466323216757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.29929065704345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.34619732697804767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.4769493341445923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.6163253386815389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,power_law_1.2,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,power_law_1.2,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,power_law_1.2,0.07301120162010193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,power_law_1.2,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,power_law_1.2,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,power_law_1.2,0.08168320059776306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,power_law_1.2,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,power_law_1.2,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,power_law_1.2,0.09370880126953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,power_law_1.2,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,power_law_1.2,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,power_law_1.2,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.14186240434646608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.15243519544601442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.2025279998779297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.21731839179992676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.3257472038269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.40008320808410647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,power_law_1.2,0.5735104084014893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,power_law_1.2,0.7166975975036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,power_law_1.2,0.04010879993438721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,power_law_1.2,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,power_law_1.2,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,power_law_1.2,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,power_law_1.2,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,power_law_1.2,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,power_law_1.2,0.08552320003509521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,power_law_1.2,0.09991040229797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,power_law_1.2,0.13377280235290528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,power_law_1.2,0.17106560468673707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,power_law_1.2,0.22663040161132814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,power_law_1.2,0.28206720352172854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,power_law_1.2,0.35701119899749756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,balanced,0.0746613343556722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,balanced,0.10207466284434001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,balanced,0.1546986699104309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,balanced,0.2576853235562642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,balanced,0.46668267250061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,balanced,0.5694293181101481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,balanced,0.5734239816665649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,balanced,0.5751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,balanced,0.5727680126825968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.2,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.2,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.2,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,balanced,0.5745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,balanced,0.5785226821899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,balanced,0.5801493326822916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,balanced,0.5823306639989217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.2,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,balanced,0.588266650835673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,balanced,0.5953066746393839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,balanced,0.5984053214391073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.2,0.07760639786720276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,balanced,0.6070026556650797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,balanced,0.6485066811243693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,balanced,0.6320906480153402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.2,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,balanced,0.7940320173899332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,balanced,0.7011040051778158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.2,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,balanced,0.8527946472167969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.2,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,balanced,0.8287466367085775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,balanced,1.3016106287638347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.2,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,balanced,1.2373600006103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.2,0.08351359963417053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,balanced,1.6783253351847331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.2,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.2,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,balanced,2.1745173136393228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.2,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.2,0.10062079429626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.2,0.11436799764633179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.2,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.2,0.13427200317382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.2,0.17063039541244507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.2,0.18398720026016235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.2,0.2422015905380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.2,0.27141120433807375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.2,0.3833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.2,0.4889855861663818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.2,0.7632448196411132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.2,0.822764778137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,power_law_1.2,0.6355904102325439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.1245695948600769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.13648639917373656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.1502784013748169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.14959360361099244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.15406080484390258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.16299519538879395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.16319359540939332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.174892795085907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.17679359912872314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.18906879425048828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.20638079643249513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.235916805267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.2856575965881348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.33244800567626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.4343552112579346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.5522816181182861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.7325056076049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,0.9356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.3160575866699218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,1.74267520904541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,power_law_1.2,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,power_law_1.2,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,power_law_1.2,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,power_law_1.2,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,power_law_1.2,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,power_law_1.2,0.13543039560317993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,power_law_1.2,0.1816383957862854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,power_law_1.2,0.06457599997520447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,power_law_1.2,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,power_law_1.2,0.06507520079612732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,power_law_1.2,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,power_law_1.2,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,power_law_1.2,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,power_law_1.2,0.07100160121917724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,power_law_1.2,0.08319360017776489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,power_law_1.2,0.08444160223007202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,power_law_1.2,0.0852288007736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,power_law_1.2,0.11206400394439697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,power_law_1.2,0.11447039842605591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,power_law_1.2,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,power_law_1.2,0.22298879623413087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,power_law_1.2,0.17308160066604614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,power_law_1.2,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,power_law_1.2,0.24277760982513427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,power_law_1.2,0.287987208366394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,power_law_1.2,0.3886528015136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,power_law_1.2,0.3148799896240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,power_law_1.2,0.5335040092468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.07810559868812561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.10925439596176148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.11325440406799317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.11564799547195434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.12040959596633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.12425600290298462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.12071039676666259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.130348801612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,power_law_1.2,0.7526016235351562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.1437440037727356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.16416000127792357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.1832128047943115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.21819519996643066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.21987199783325195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.251692795753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,power_law_1.2,0.33402879238128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,power_law_1.2,0.3872895956039429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,power_law_1.2,0.4542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,power_law_1.2,0.5383103847503662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,power_law_1.2,0.9571455955505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,power_law_1.2,0.5716032028198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,power_law_1.2,0.6848383903503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,power_law_1.2,0.9080191612243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,power_law_1.2,1.464409637451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,power_law_1.2,0.7503935813903808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,power_law_1.2,1.8794879913330078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.08977919816970825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.09828479886054993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.11690880060195923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.17533440589904786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.20935680866241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.2387903928756714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3160640001296997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.3726975917816162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.5171135902404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.15912959575653077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.21573119163513182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.3210943937301636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.42938880920410155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.43009281158447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.4381375789642334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.4392064094543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.43997440338134763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.4400320053100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.4389823913574219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.6506752014160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.4447167873382568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.4524864196777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.4556735992431641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.4783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.5172927856445313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5265408039093018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.5761087894439697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.6751552104949952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,0.7945280075073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.0046272277832031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.2235775947570802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,1.6993728637695313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,0.9242815971374512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,2.1648256301879885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.2084351539611817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,3.1397056579589844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,4.065222549438476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,5.995980834960937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,7.895763397216797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,power_law_1.2,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,power_law_1.2,0.02953599989414215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,power_law_1.2,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,power_law_1.2,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,power_law_1.2,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,power_law_1.2,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,power_law_1.2,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,power_law_1.2,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,power_law_1.2,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,power_law_1.2,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,power_law_1.2,0.06929280161857605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,power_law_1.2,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,power_law_1.2,0.11499520540237426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,power_law_1.2,0.15569920539855958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,power_law_1.2,0.18490240573883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,power_law_1.2,0.26245119571685793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,power_law_1.2,0.337228798866272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,power_law_1.2,0.4912447929382324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,power_law_1.2,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,power_law_1.2,0.10094079971313477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,power_law_1.2,0.1335744023323059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,power_law_1.2,0.14715520143508912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,power_law_1.2,0.6434175968170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,power_law_1.2,0.19652479887008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,power_law_1.2,0.208076810836792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,power_law_1.2,0.25001599788665774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,power_law_1.2,0.2710720062255859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,power_law_1.2,0.3049407958984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,power_law_1.2,0.3083775997161865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,power_law_1.2,0.2739264011383057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,power_law_1.2,0.31598079204559326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,power_law_1.2,0.30979199409484864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,power_law_1.2,0.3031424045562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,power_law_1.2,0.30122239589691163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,power_law_1.2,0.3456000089645386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,power_law_1.2,0.3429824113845825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,power_law_1.2,0.3704063892364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,power_law_1.2,0.4131135940551758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,power_law_1.2,0.4967487812042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,power_law_1.2,0.5145023822784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,power_law_1.2,0.6525504112243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,power_law_1.2,0.752345609664917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,power_law_1.2,0.9944128036499024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,power_law_1.2,1.1494848251342773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,power_law_1.2,1.5178239822387696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,power_law_1.2,2.0048831939697265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.2,0.12123520374298095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.2,0.1466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.2,0.14684159755706788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.2,0.1422144055366516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.2,0.1425984025001526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.2,0.14339200258255005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.2,0.1508095979690552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.2,0.14525439739227294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.2,0.1444991946220398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.2,0.1557952046394348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.2,0.16003199815750122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.2,0.15783040523529052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.2,0.16768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.2,0.17114880084991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.2,0.1876031994819641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.2,0.21704320907592772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.2,0.2281343936920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.2,0.2769792079925537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.2,0.3225343942642212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.2,0.42008957862854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.2,0.41205759048461915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.2,0.5451327800750733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.2,0.7050943851470948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.2,1.0050944328308105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.2,1.1088319778442384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,balanced,0.05406400064627329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,balanced,0.07904533545176189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,balanced,0.0777759999036789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,balanced,0.07964799801508586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,balanced,0.08200533191363017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,balanced,0.08507200082143147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,balanced,0.08654933174451192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,balanced,0.08794666330019633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,balanced,0.09416533509890239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,balanced,0.10237333178520203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,balanced,0.1074133316675822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,balanced,0.10703466335932414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,balanced,0.131221334139506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,balanced,0.13777599732081094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,balanced,0.1820533275604248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,balanced,0.22113599379857382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,balanced,0.29307732979456586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,balanced,0.345578670501709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,balanced,0.49164799849192303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,balanced,0.6151626507441202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,balanced,0.8923146724700928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,balanced,1.159658670425415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.05076479911804199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.0759168028831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.0673919975757599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.0630016028881073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.09251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.11313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.0768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.1366976022720337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.18629759550094604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.17888640165328978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.26693120002746584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.3428800106048584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.08826879858970642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.47199358940124514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.6821055889129639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.1362239956855774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.15464320182800292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.20250239372253417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.23760640621185303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.31947519779205324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.3832639932632446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.5721151828765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,0.809932804107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.09059839844703674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.08195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.08447359800338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.08741120100021363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.09751039743423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.10084480047225952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.10836479663848878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.14954880475997925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.2031615972518921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.25136001110076905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.3524928092956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.4614528179168701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.6710015773773194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.7988800048828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.07733759880065919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.07210879921913146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.07316480278968811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.10828800201416015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.1212928056716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.1489408016204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.16300159692764282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.20394880771636964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.23824000358581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.3432512044906616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.4204224109649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,power_law_1.2,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,power_law_1.2,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,power_law_1.2,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,power_law_1.2,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,power_law_1.2,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,power_law_1.2,0.10625280141830444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,power_law_1.2,0.10641280412673951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,power_law_1.2,0.11147520542144776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,power_law_1.2,0.11073919534683227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,power_law_1.2,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,power_law_1.2,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.677888011932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,power_law_1.2,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,power_law_1.2,0.12399359941482543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,power_law_1.2,0.12572799921035765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,power_law_1.2,0.13370239734649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,power_law_1.2,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.2,0.1805184006690979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.2,0.22678399085998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.2,0.27134079933166505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.2,0.3874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.2,0.4338047981262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.2,0.629702377319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.2,0.8675264358520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.2,1.3285696029663085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.2,1.6285055160522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.8986047744750977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,balanced,0.029445332785447437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,balanced,0.02941333254178365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,balanced,0.030218665798505146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,balanced,0.029189333319664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,balanced,0.03410666684309641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,balanced,0.03919466584920883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,balanced,0.08130666613578796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,balanced,0.1153546671072642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,balanced,0.14467199643452963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,balanced,0.08029333253701527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,balanced,0.11187733213106792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,balanced,0.17967466513315836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,balanced,0.3035893241564433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,balanced,0.3057066599527995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,balanced,0.3086346586545308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,balanced,0.3107733329137166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,balanced,0.3113173246383667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,balanced,0.31218133370081586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,balanced,0.3153866728146871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,balanced,0.3155306577682495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,balanced,0.32200533151626587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,balanced,0.3266293406486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,balanced,0.3328160047531128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,balanced,0.34177064895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,balanced,0.3587253491083781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,balanced,0.3744746843973796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,balanced,0.4091200033823649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,balanced,0.4503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,balanced,0.5291306575139364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,balanced,0.5794133345286051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,balanced,0.7634666760762533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,balanced,0.8612639904022217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,balanced,1.2301599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,balanced,1.4897759755452473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.08439040184020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.08894079923629761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.11260160207748413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.12315520048141479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.14317439794540404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.22064640522003173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.2753024101257324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.36087679862976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.4477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.6607744216918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,power_law_1.01,0.8251008033752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.06222079992294312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.09096320271492005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.10149120092391968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.14344320297241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.18288639783859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.2300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.2969599962234497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.3899967908859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.5517183780670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.7690048217773438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,power_law_1.2,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,power_law_1.2,0.13793280124664306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,power_law_1.2,0.1804927945137024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,power_law_1.2,0.2518656015396118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,power_law_1.2,0.3142591953277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,power_law_1.2,0.355295991897583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,power_law_1.2,0.4833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,power_law_1.2,0.4841919898986816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,power_law_1.2,0.46682238578796387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,power_law_1.2,0.5119679927825928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,power_law_1.2,0.5213696002960205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,power_law_1.2,0.5506559848785401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,power_law_1.2,0.5260543823242188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,power_law_1.2,0.5868415832519531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,power_law_1.2,0.6087296009063721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,power_law_1.2,0.6159167766571045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,power_law_1.2,0.6592576026916503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,power_law_1.2,0.7052031993865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,power_law_1.2,0.8006655693054199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,power_law_1.2,0.8933055877685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,power_law_1.2,1.0864895820617675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,power_law_1.2,1.2295040130615233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,power_law_1.2,1.5799807548522948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,power_law_1.2,1.863488006591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,power_law_1.2,2.4758783340454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,power_law_1.2,2.860063934326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,balanced,0.09756799538930257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,balanced,0.1476640005906423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,balanced,0.24242132902145386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,balanced,0.43818668524424237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,balanced,0.8257439931233724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,balanced,1.6027040481567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,balanced,1.606218655904134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,balanced,1.6136479377746582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,balanced,1.6125226020812988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,balanced,1.6224479675292969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,balanced,1.6212266286214192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,balanced,1.6310133934020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,balanced,1.6386772791544597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,balanced,1.6473066012064617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,balanced,1.6614880561828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,balanced,1.6685226758321126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,balanced,1.693402608235677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,balanced,1.7351466814676921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,balanced,1.7977387110392253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,balanced,1.8649013837178547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,balanced,1.9496533075968425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,balanced,2.089738686879476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,balanced,2.266554673512777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,balanced,2.7577012379964194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,balanced,2.8348960876464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,balanced,4.324207941691081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,balanced,4.544165293375651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.14860160350799562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.2343616008758545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.230348801612854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.31848320960998533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.3403968095779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.3242815971374512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.38893439769744875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.34152960777282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.41239681243896487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.39426560401916505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.3747391939163208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.3603584051132202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.3999167919158936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.3980479955673218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.4215871810913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.4327807903289795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.4720191955566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.5311615943908692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.5279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,power_law_1.2,0.6493887901306152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,power_law_1.2,0.5893439769744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,power_law_1.2,0.7029312133789063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,power_law_1.2,0.7807231903076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,power_law_1.2,0.9607487678527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,power_law_1.2,1.1995327949523926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,power_law_1.2,1.8916543960571288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,power_law_1.2,2.1876224517822265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.06235520243644714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.1325503945350647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.1769600033760071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.20265600681304932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.27346560955047605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.34357120990753176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.4934591770172119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.622489595413208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.05612159967422485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.13583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.16344959735870362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.20886399745941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.21459200382232665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.21419520378112794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.21755518913269042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.2185663938522339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.23304319381713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.2381376028060913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.25601921081542967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.26538240909576416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.2708415985107422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.2878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.3150912046432495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.33857920169830324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.40894079208374023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.4612736225128174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.5781248092651368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,power_law_1.2,0.689241600036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,power_law_1.2,0.9309311866760254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.0780608177185058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,power_law_1.2,1.4725376129150392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,power_law_1.2,1.8318464279174804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,balanced,0.07705066601435344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,balanced,0.10424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,balanced,0.1183093289534251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,balanced,0.11764267086982727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,balanced,0.11786133050918579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,balanced,0.11939199765523274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,balanced,0.1209386686484019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,balanced,0.12127466996510823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,balanced,0.12147733569145203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,balanced,0.12390933434168498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,balanced,0.12596799929936728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,balanced,0.13089600205421448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,balanced,0.1371946632862091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,balanced,0.1339413324991862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,balanced,0.1511626640955607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,balanced,0.15306666493415833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,balanced,0.18118399381637573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,balanced,0.18508267402648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,balanced,0.24040534098943075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,balanced,0.2505173285802205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,balanced,0.38309331734975177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,balanced,0.4071466525395711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,balanced,0.5911680062611898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,balanced,0.7068266868591309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,power_law_1.01,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,power_law_1.01,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,power_law_1.01,0.08384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,power_law_1.01,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,power_law_1.01,0.12458239793777466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,power_law_1.01,0.16362240314483642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,power_law_1.01,0.1701311945915222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,power_law_1.01,0.16554239988327027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,power_law_1.01,0.17345919609069824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,power_law_1.01,0.17374720573425292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,power_law_1.01,0.18033280372619628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,power_law_1.01,0.1857792019844055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,power_law_1.01,0.19304959774017333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,power_law_1.01,0.19555840492248536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,power_law_1.01,0.21156480312347412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,power_law_1.01,0.2166912078857422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,power_law_1.01,0.2283008098602295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,power_law_1.01,0.2530751943588257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,power_law_1.01,0.2915136098861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,power_law_1.01,0.34618239402770995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,power_law_1.01,0.4215871810913086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,power_law_1.01,0.545363187789917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,power_law_1.01,0.6049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,power_law_1.01,0.8272768020629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,power_law_1.01,1.120729637145996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,power_law_1.01,1.5441408157348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,power_law_1.01,1.8682815551757812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.01,0.028575998544692994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.01,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.01,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.01,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.01,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.01,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.01,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.01,0.13360639810562133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.01,0.18630399703979492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.01,0.23105919361114502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,balanced,0.07238933444023132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,balanced,0.07237866520881653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,balanced,0.07289066910743713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,balanced,0.0757066657145818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,balanced,0.07361066838105519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,balanced,0.07770666480064392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,balanced,0.07707199951012929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,balanced,0.08160533507664998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,balanced,0.08354133367538452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,balanced,0.08906666437784831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,balanced,0.09316800038019817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,balanced,0.10389332969983418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,balanced,0.1116480032602946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,balanced,0.12934399644533792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,balanced,0.14882666865984598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,balanced,0.1885813275973002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,balanced,0.2218773365020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,balanced,0.30458666880925495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,balanced,0.37586132685343426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,power_law_1.2,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,power_law_1.2,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,power_law_1.2,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,power_law_1.2,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.2,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.2,0.11543040275573731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.03695360124111176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.10120320320129395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.1177024006843567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.2,0.1324928045272827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.1478592038154602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.2,0.1705471992492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.177183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.20037760734558105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.2,0.24086399078369142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.29624319076538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.42842879295349123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.5264063835144043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.2,0.3108288049697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.2,0.4196608066558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.2,0.1303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.2,0.15470080375671386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.2,0.19039360284805298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.2,0.20595200061798097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.2,0.2930815935134888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.2,0.2812544107437134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.2,0.2826047897338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.2,0.29278719425201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.2,0.29684479236602784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.2,0.30925440788269043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.2,0.31738240718841554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.2,0.32039680480957033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.2,0.34081919193267823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.2,0.35306239128112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.2,0.35294721126556394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.2,0.3707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.2,0.4296256065368652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.2,0.47735037803649905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.2,0.5660927772521973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.2,0.6664959907531738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.2,0.8497471809387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.2,0.9847359657287598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.2,1.245913600921631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.2,1.6731136322021485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.2,2.1599679946899415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.2,2.874444770812988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.2,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.2,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.2,0.10571520328521729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.2,0.11441919803619385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.2,0.11802239418029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.2,0.12174079418182374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.2,0.12341760396957398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.2,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.2,0.12837120294570922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.2,0.1310912013053894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.2,0.13436800241470337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.2,0.1440127968788147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.2,0.15473920106887817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.2,0.17416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.2,0.17816319465637206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.2,0.20904960632324218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.2,0.23775360584259034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.2,0.29100799560546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.2,0.3452608108520508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.2,0.447327995300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.2,0.5571199893951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.2,0.7727231979370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.2,0.9854592323303223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,balanced,0.033071999748547874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,balanced,0.03581333408753077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,balanced,0.0454773356517156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,balanced,0.05983466903368632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,balanced,0.0662613312403361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,balanced,0.07912000020345052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,balanced,0.10121066371599834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,balanced,0.12359467148780823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,balanced,0.17379732926686606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,balanced,0.21059733629226685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,balanced,0.3006613254547119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,balanced,0.40666667620340985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.04283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.04000000059604645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.04433279931545257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.12295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.16890239715576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.19788800477981566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.2981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.3623807907104492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.5090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.028492799401283263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.02839039862155914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.02951039969921112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.06079999804496765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.1575551986694336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.23464319705963135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.3049407958984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.09151359796524047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.09279360175132752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.09380480051040649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.1060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.11612800359725953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.13818880319595336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.15905280113220216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.18706560134887695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.2342207908630371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.295795202255249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.3855488061904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.46187520027160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,power_law_1.01,0.6779136180877685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,power_law_1.01,0.8477888107299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,power_law_1.2,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,power_law_1.2,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,power_law_1.2,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,power_law_1.2,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,power_law_1.2,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,power_law_1.2,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,power_law_1.2,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,power_law_1.2,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,power_law_1.2,0.0921280026435852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,power_law_1.2,0.1066815972328186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,power_law_1.2,0.15239039659500123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,power_law_1.2,0.17895679473876952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,power_law_1.2,0.2582848072052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,power_law_1.2,0.3672895908355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,power_law_1.2,0.4291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,power_law_1.2,0.6307136058807373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,power_law_1.2,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,power_law_1.2,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.06874880194664001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.07936639785766601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.12513920068740844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.18710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.20725760459899903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.3172991991043091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.15971839427947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.19831680059432982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.19506560564041137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.19466240406036378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.1742784023284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.18152960538864135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.16442879438400268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.17689599990844726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.17631360292434692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.1691264033317566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.18007680177688598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.17216639518737792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.20007040500640869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.19698560237884521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.2366015911102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.24591999053955077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.3361920118331909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.32213759422302246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.40331521034240725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,power_law_1.2,0.4459968090057373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,power_law_1.2,0.5958847999572754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,power_law_1.2,0.6265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,power_law_1.2,0.9235584259033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,power_law_1.2,1.1728896141052245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.2119231939315796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.38524799346923827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.08307200074195861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.09128959774971009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.10313600301742554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.12645119428634644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.14625279903411864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.18570879697799683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.21281919479370118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.27745280265808103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.34586880207061765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.5048575878143311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.658406400680542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,power_law_1.01,0.05514879822731018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,power_law_1.01,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,power_law_1.01,0.08447359800338745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,power_law_1.01,0.08616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,power_law_1.01,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,power_law_1.01,0.08834559917449951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,power_law_1.01,0.08651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,power_law_1.01,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,power_law_1.01,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,power_law_1.01,0.10290559530258178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,power_law_1.01,0.1060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,power_law_1.01,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,power_law_1.01,0.12613120079040527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,power_law_1.01,0.1410048007965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,power_law_1.01,0.16940159797668458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,power_law_1.01,0.1926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,power_law_1.01,0.24384000301361083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,power_law_1.01,0.2945535898208618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,power_law_1.01,0.4204927921295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,power_law_1.01,0.5048575878143311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,power_law_1.01,0.7335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,power_law_1.01,0.9430335998535156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,balanced,0.04550399879614512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,balanced,0.050581331054369606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,balanced,0.0605973352988561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,balanced,0.06418133278687795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,balanced,0.08525333801905315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,balanced,0.10355200370152791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,balanced,0.14269333084424338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,balanced,0.16565333803494772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,balanced,0.22136000792185465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,balanced,0.26472000281016034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,balanced,0.3720853328704834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,balanced,0.47195200125376385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.06687359809875489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.0769599974155426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.1355455994606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.17637759447097778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.21144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.2775167942047119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.3538431882858276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,power_law_1.2,0.4998784065246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,power_law_1.01,0.02635520100593567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,power_law_1.01,0.02513279914855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,power_law_1.01,0.024403199553489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,power_law_1.01,0.023756800591945647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,power_law_1.01,0.022918400168418885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,power_law_1.01,0.02650879919528961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,power_law_1.01,0.026572799682617186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,power_law_1.01,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,power_law_1.01,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,power_law_1.01,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.06342399716377259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.0768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.10037120580673217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.1727552056312561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,power_law_1.2,0.632147216796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.19724160432815552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,power_law_1.01,0.09790719747543335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,power_law_1.01,0.11464320421218872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,power_law_1.01,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,power_law_1.01,0.1283903956413269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,power_law_1.01,0.11054079532623291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,power_law_1.01,0.1404096007347107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,power_law_1.01,0.13612799644470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,power_law_1.01,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,power_law_1.01,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,power_law_1.01,0.14362239837646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,power_law_1.01,0.13793280124664306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,power_law_1.01,0.14780800342559813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,power_law_1.01,0.1549631953239441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,power_law_1.01,0.17247999906539918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,power_law_1.01,0.1739456057548523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,power_law_1.01,0.1979647994041443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,power_law_1.01,0.2079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,power_law_1.01,0.2405695915222168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,power_law_1.01,0.2848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,power_law_1.01,0.3500735998153687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,power_law_1.01,0.3947200059890747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,power_law_1.01,0.5363776206970214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,power_law_1.01,0.7060671806335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,power_law_1.01,0.9470784187316894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,power_law_1.01,1.1940159797668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,balanced,0.034634667138258614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,balanced,0.041146665811538696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,balanced,0.07323200007279713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,balanced,0.10381333033243816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,balanced,0.130213330189387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,balanced,0.17140267292658487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,balanced,0.18897066513697305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,balanced,0.2651413281758626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,balanced,0.343509316444397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,balanced,0.4904373486836751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,balanced,0.6227786540985107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,balanced,0.13858133554458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,balanced,0.23642132679621378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,balanced,0.2386773427327474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,balanced,0.23817066351572672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,balanced,0.24009599288304648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,balanced,0.24200000365575156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,balanced,0.24421334266662598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,balanced,0.24478934208552042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,balanced,0.246453324953715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,balanced,0.2505386670430501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,balanced,0.2635999917984009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,balanced,0.26412800947825116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,balanced,0.26870399713516235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,balanced,0.2890666723251343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,balanced,0.29230932394663495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,balanced,0.3014986713727315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,balanced,0.37015998363494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,balanced,0.37460800011952716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,balanced,0.5089973211288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,balanced,0.5632693370183309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,balanced,0.7783040205637614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,balanced,0.9587306976318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,balanced,1.4067573547363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,balanced,1.839626630147298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,balanced,2.720997174580892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,balanced,3.5695679982503257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,power_law_1.01,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,power_law_1.01,0.15062400102615356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,power_law_1.01,0.14677759408950805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,power_law_1.01,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,power_law_1.01,0.1347264051437378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,power_law_1.01,0.13765759468078614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,power_law_1.01,0.1342144012451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,power_law_1.01,0.1401471972465515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,power_law_1.01,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,power_law_1.01,0.14079359769821168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,power_law_1.01,0.14540159702301025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,power_law_1.01,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,power_law_1.01,0.151910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,power_law_1.01,0.15955840349197387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,power_law_1.01,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,power_law_1.01,0.17437440156936646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,power_law_1.01,0.18730239868164061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,power_law_1.01,0.20234239101409912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,power_law_1.01,0.2333375930786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,power_law_1.01,0.25868160724639894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,power_law_1.01,0.31192960739135744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,power_law_1.01,0.39134719371795657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,power_law_1.01,0.5141823768615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,power_law_1.01,0.6213568210601806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,power_law_1.01,0.8712767601013184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,power_law_1.01,1.0703488349914552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.06853119730949402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.07836160063743591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.09534080028533935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.1020095944404602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.11989760398864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.15104000568389891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.1716096043586731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.21895039081573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.24893438816070557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.33201920986175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,power_law_1.01,0.39429759979248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,power_law_1.01,0.5612287998199463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,power_law_1.01,0.7286975860595704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,balanced,0.06387733419736226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,balanced,0.06273599962393443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,balanced,0.06746133168538411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,balanced,0.06481599807739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,balanced,0.06705600023269653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,balanced,0.0675786683956782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,balanced,0.06610133250554402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,balanced,0.06710400183995564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,balanced,0.06829333305358887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,balanced,0.06956799825032552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,balanced,0.07403199871381123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,balanced,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,balanced,0.0842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,balanced,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,balanced,0.09608532985051473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,balanced,0.13301333785057068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,balanced,0.13050666451454163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,balanced,0.17787732680638632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,balanced,0.20378132661183676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,balanced,0.27980266014734906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,balanced,0.34622399012247723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,balanced,0.49724264939626056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,balanced,0.6423466602961222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,balanced,0.06458133459091187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,balanced,0.08260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,balanced,0.08949866890907288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,balanced,0.08901866277058919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,balanced,0.09124267101287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,balanced,0.0904319981733958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,balanced,0.0930560032526652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,balanced,0.09381866455078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,balanced,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,balanced,0.10036800305048625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,balanced,0.10571199655532837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,balanced,0.12960533301035562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,balanced,0.13784000277519226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,balanced,0.1732906699180603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,balanced,0.20594133933385214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,balanced,0.2848586638768514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,balanced,0.36394135157267254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,balanced,0.524069348971049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,balanced,0.6805333296457926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,balanced,1.0030240217844646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,balanced,1.2995413144429524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.01,0.08231040239334106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.01,0.09827200174331666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.01,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.01,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.01,0.1023743987083435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.01,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.01,0.10467840433120727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.01,0.10432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.01,0.10401920080184937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.01,0.10969599485397338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.01,0.10936959981918334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.01,0.11191040277481079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.01,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.01,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.01,0.16295679807662963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.01,0.1625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.01,0.19982080459594725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.01,0.24447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.01,0.3171456098556519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.01,0.3479487895965576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.01,0.47173118591308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.01,0.5528255939483643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.01,0.7764031887054443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.01,1.0258048057556153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.2,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.2,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.2,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.2,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.2,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.2,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.2,0.07812479734420777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.2,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.2,0.08543360233306885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.2,0.08475520014762879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.2,0.10289280414581299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.2,0.10589439868927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.2,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.2,0.1352959990501404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.2,0.16412160396575928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.2,0.19445120096206664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.2,0.25046401023864745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.2,0.27616000175476074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.2,0.35778560638427737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.2,0.4193984031677246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.2,0.6479743957519531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.2,0.809721565246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,balanced,0.0734986662864685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,balanced,0.10188266634941101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,balanced,0.15134933590888977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,balanced,0.15213867028554282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,balanced,0.15779733657836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,balanced,0.15898666779200235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,balanced,0.15733333428700766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,balanced,0.15970133741696677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,balanced,0.1609173317750295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,balanced,0.16181866327921549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,balanced,0.16332800189654031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,balanced,0.17037334044774374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,balanced,0.17246933778127035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,balanced,0.1797706683476766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,balanced,0.19309866428375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,balanced,0.20774932702382407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,balanced,0.2311519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,balanced,0.25491732358932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,balanced,0.31330132484436035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,balanced,0.36087465286254883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,balanced,0.4706079959869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,balanced,0.5567946831385294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,balanced,0.781984011332194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,balanced,0.9833813508351644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.058790397644042966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.08725759983062745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.09981439709663391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.11199359893798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.12317440509796143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.13949439525604249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.15894399881362914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.18231680393218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.29111039638519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.35313279628753663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.44686079025268555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.5883840084075928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.8069696426391602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,0.9475263595581055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.1202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.13619199991226197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.19040000438690186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,power_law_1.2,0.23955199718475342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,power_law_1.2,0.33547520637512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,power_law_1.2,0.5262976169586182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.08223999738693237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.10090880393981934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.10346239805221558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.12341120243072509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.1342911958694458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.1700096011161804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.20471038818359374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.27682559490203856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.32258560657501223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,power_law_1.2,0.49111042022705076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,power_law_1.2,0.0383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,power_law_1.2,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,power_law_1.2,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,power_law_1.2,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,power_law_1.2,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,power_law_1.2,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,power_law_1.2,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,power_law_1.2,0.09041280150413514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,power_law_1.2,0.1283776044845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.2,0.1457856059074402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.2,0.1991104006767273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,power_law_1.2,0.5554368019104003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.2,0.24423680305480958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.2,0.3682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.2,0.5093311786651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,power_law_1.2,0.9848832130432129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.2,0.7552127838134766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.2,0.9639679908752441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,power_law_1.2,1.0444416046142577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.2,1.4258175849914552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.2,1.939641571044922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.06117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.115174400806427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.13790080547332764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.1256384015083313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.17529599666595458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,power_law_1.2,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.19107199907302858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.2588160037994385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,power_law_1.01,0.31499519348144533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,power_law_1.01,0.4539648056030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,power_law_1.2,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,power_law_1.01,0.5916607856750489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,power_law_1.2,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,power_law_1.2,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,power_law_1.2,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,power_law_1.2,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.2,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.2,0.13984639644622804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.2,0.1663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.2,0.24206719398498536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.2,0.28813440799713136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.2,0.47283201217651366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.2,0.6020607948303223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.2,0.9004991531372071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.2,1.183199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,balanced,0.0505920002857844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,balanced,0.051589335004488625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,balanced,0.08338666955629985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,balanced,0.11213866869608562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,balanced,0.11515733599662781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,balanced,0.11595199505488078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,balanced,0.1160586675008138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,balanced,0.11707733074824016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,balanced,0.11725866794586182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,balanced,0.1188320020834605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,balanced,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,balanced,0.12011200189590454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,balanced,0.12662933270136514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,balanced,0.12776000301043192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,balanced,0.12991467118263245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,balanced,0.14019733667373657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.14499200383822122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.16038933396339417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.17753066619237265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.20388267437616983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.23444799582163492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.29892265796661377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.353274663289388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.48476799329121906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.5797386566797892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,balanced,0.07011199990908305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,balanced,0.08501332998275757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,balanced,0.1228000024954478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,balanced,0.18312533696492514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,balanced,0.185370663801829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,balanced,0.18177066246668497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,balanced,0.1821920077006022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,balanced,0.1848479906717936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,balanced,0.1858560045560201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,balanced,0.18821332852045694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,balanced,0.19131199518839517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,balanced,0.19376534223556519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,balanced,0.2009920080502828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,balanced,0.20453866322835287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,balanced,0.23360532522201538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.251525342464447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.29149333635965985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.3248800039291382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.41413335005442303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.49855466683705646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.7119786739349365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.8838880062103271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,balanced,1.2765066623687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,balanced,1.6417226791381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.01,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.01,0.0901632010936737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.01,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.01,0.1055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.01,0.12368639707565307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.01,0.1483199954032898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.01,0.14118399620056152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.01,0.14103039503097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.01,0.14455679655075074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.01,0.15338879823684692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.01,0.15408639907836913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.01,0.1532096028327942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.01,0.15457279682159425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.01,0.1633407950401306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.01,0.17276159524917603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.01,0.17916159629821776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.01,0.19371520280838012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.01,0.22041599750518798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.01,0.24333438873291016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.01,0.29675519466400146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.01,0.3561408042907715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.01,0.4546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.01,0.548089599609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.01,0.7464767932891846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.01,0.8713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.01,1.2256832122802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.01,1.6135679244995118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.11038719415664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.16032639741897584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.23257598876953126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.25798399448394777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,power_law_1.2,0.40309758186340333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,balanced,0.07820266485214233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,balanced,0.12012799580891927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,balanced,0.20323733488718668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,balanced,0.20361065864562988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,balanced,0.20165866613388062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,balanced,0.20480533440907797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,balanced,0.20808533827463785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,balanced,0.20708799362182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,balanced,0.20592532555262247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,balanced,0.2066239913304647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,balanced,0.2075093388557434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,balanced,0.2111626664797465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,balanced,0.21729065974553427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,balanced,0.21712533632914224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,balanced,0.22236265738805136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,balanced,0.22822399934132895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,balanced,0.24357332785924277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,balanced,0.26975466807683307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,balanced,0.26981866359710693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,balanced,0.3254186709721883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,balanced,0.3471946716308594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,balanced,0.4355253378550212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,balanced,0.4960319995880127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,balanced,0.6633866628011068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,balanced,0.844101349512736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,power_law_1.2,0.48732800483703614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.05076479911804199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.09377920031547546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.12318079471588135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.14553600549697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.15965440273284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.20366079807281495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.2667840003967285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.33739519119262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,power_law_1.2,0.4394495964050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,power_law_1.2,0.6430272102355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,power_law_1.2,0.6882624149322509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.061478400230407716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.1252351999282837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.1696447968482971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.2366015911102295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,balanced,0.06493866443634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,balanced,0.09186666210492452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,balanced,0.14350400368372598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,balanced,0.24286399284998575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,balanced,0.43745601177215576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,balanced,0.4368373155593872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,balanced,0.4389066696166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,balanced,0.4436746835708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,balanced,0.44340264797210693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,balanced,0.4413546721140544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,balanced,0.4442773262659709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,balanced,0.4479946692784627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,balanced,0.45026131470998126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,balanced,0.45535465081532794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,balanced,0.4569386641184489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,balanced,0.46009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,balanced,0.4695306619008382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,balanced,0.48337066173553467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,balanced,0.4995946486790975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,balanced,0.527237335840861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,balanced,0.559936006863912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,balanced,0.6173706849416097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,balanced,0.6934826374053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,balanced,0.7495893637339274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,balanced,1.0212799708048503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,balanced,1.1546293099721272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.33863680362701415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,power_law_1.2,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,power_law_1.2,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,power_law_1.2,0.07790719866752624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,power_law_1.2,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,power_law_1.2,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,power_law_1.2,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,power_law_1.2,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,power_law_1.2,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,power_law_1.2,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,power_law_1.2,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,power_law_1.2,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,power_law_1.2,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,power_law_1.2,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,power_law_1.2,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,power_law_1.2,0.11161600351333618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,power_law_1.2,0.11152640581130982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,power_law_1.2,0.13906559944152833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,power_law_1.2,0.1534783959388733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,power_law_1.2,0.1932031989097595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,power_law_1.2,0.23440001010894776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,power_law_1.2,0.29608960151672364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,power_law_1.2,0.3671168088912964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,power_law_1.2,0.5920191764831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,power_law_1.2,0.7210559844970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,power_law_1.01,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,power_law_1.01,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,power_law_1.01,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,power_law_1.01,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,power_law_1.01,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,power_law_1.01,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,power_law_1.01,0.10732799768447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,power_law_1.01,0.15544320344924928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,power_law_1.01,0.1909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,power_law_1.01,0.29119999408721925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,power_law_1.01,0.35104639530181886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,balanced,0.10299733281135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,balanced,0.15159466862678528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,balanced,0.2490666707356771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,balanced,0.44540266195933026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,balanced,0.8415466944376627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,balanced,1.2336266835530598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,balanced,1.2370826403299968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,balanced,1.2393333117167156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,balanced,1.2414453029632568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,balanced,1.249237298965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,balanced,1.2604906558990479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,balanced,1.2631253401438396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,balanced,1.2743733723958333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,balanced,1.289072036743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,balanced,1.300101359685262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,balanced,1.3301066557566326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,balanced,1.3781545956929524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,balanced,1.4242666562398274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,balanced,1.5213120778401692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,balanced,1.5825173060099285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,balanced,1.752346674601237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,balanced,1.9227412541707356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,balanced,2.2515199979146323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,balanced,2.7083520889282227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,balanced,3.3729120890299478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,balanced,4.495200157165527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,power_law_1.01,0.5316224098205566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,power_law_1.01,0.6280447959899902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,power_law_1.2,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,power_law_1.2,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,power_law_1.2,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,power_law_1.2,0.07317759990692138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,power_law_1.2,0.08501759767532349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,power_law_1.2,0.08718079924583436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,power_law_1.2,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,power_law_1.2,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,power_law_1.2,0.11861759424209595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,power_law_1.2,0.12572799921035765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,power_law_1.2,0.1474176049232483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,power_law_1.2,0.16783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,power_law_1.2,0.20684800148010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,power_law_1.2,0.23972480297088622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,power_law_1.2,0.3121151924133301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,power_law_1.2,0.3906368017196655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,power_law_1.2,0.5341631889343261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,power_law_1.2,0.674393606185913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,power_law_1.2,0.97390718460083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,power_law_1.2,1.2883008003234864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,power_law_1.2,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,power_law_1.2,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,power_law_1.2,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,power_law_1.2,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,power_law_1.2,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,power_law_1.2,0.21580801010131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,power_law_1.2,0.24099199771881102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,power_law_1.2,0.41546239852905276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,power_law_1.2,0.44315519332885744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.13375999927520751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.17461760044097902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.189900803565979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.29338879585266114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.30628480911254885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.5480576038360596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.6272768020629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,0.8646207809448242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.083308792114258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,power_law_1.01,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,power_law_1.01,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,power_law_1.01,0.08716800212860107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,power_law_1.01,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,power_law_1.01,0.10991359949111938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,power_law_1.01,0.09966719746589661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,power_law_1.01,0.11166720390319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,power_law_1.01,0.10512640476226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,power_law_1.01,0.1080064058303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,power_law_1.01,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,power_law_1.01,0.11393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,power_law_1.01,0.11662080287933349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,power_law_1.01,0.12321280241012574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,power_law_1.01,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,power_law_1.01,0.13311359882354737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,power_law_1.01,0.15084799528121948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,power_law_1.01,0.16033920049667358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,power_law_1.01,0.19023360013961793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,power_law_1.01,0.210150408744812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,power_law_1.01,0.271776008605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,power_law_1.01,0.3052095890045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,power_law_1.01,0.43184638023376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,power_law_1.01,0.5412479877471924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,power_law_1.01,0.7493631839752197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,power_law_1.01,0.9524671554565429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,balanced,0.05037866532802582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,balanced,0.051642666260401406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,balanced,0.060122668743133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,balanced,0.0792746643225352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,balanced,0.08110400040944417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,balanced,0.08080533146858215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,balanced,0.08311466872692108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,balanced,0.08417600393295288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,balanced,0.08507200082143147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,balanced,0.08505599697430928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,balanced,0.09338666995366414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,balanced,0.09217600027720134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,balanced,0.09727999567985535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,balanced,0.10646933317184448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.11179733276367188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.12794133027394614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.14123200376828512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.1690346598625183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.19962666432062784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.25596799453099567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.31278399626413983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.42748268445332843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.5423146486282349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,balanced,0.09339732925097148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,balanced,0.11331199606259663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,balanced,0.133050670226415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,balanced,0.134661336739858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,balanced,0.13329600294431052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,balanced,0.13365333278973898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,balanced,0.13545067111651102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,balanced,0.14019733667373657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,balanced,0.14059733351071677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,balanced,0.14293866356213888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,balanced,0.14469866951306662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,balanced,0.15218666195869446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,balanced,0.15505066514015198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,balanced,0.16013866662979126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,balanced,0.18027732769648233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,balanced,0.19633066654205322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,balanced,0.2731413245201111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,balanced,0.26952532927195233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,balanced,0.3732159932454427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,balanced,0.3873600165049235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,balanced,0.5483359893163046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,balanced,0.6331573327382406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,balanced,0.9206240177154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,balanced,1.1226507027943928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,balanced,0.037258667250474296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,balanced,0.039077334105968475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,balanced,0.04700266818205515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.05877333382765452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.061568001906077065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.07608533402283986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.09564266602198283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.13409599661827087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.14654933412869772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.1971786618232727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.23870933055877686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.2,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.2,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.2,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.2,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.2,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.2,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.2,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.2,0.04647040069103241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.2,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.2,0.13774720430374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.2,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.2,0.23215360641479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.2,0.28702080249786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,power_law_1.2,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,power_law_1.2,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,power_law_1.2,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,power_law_1.2,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,power_law_1.2,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,power_law_1.2,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,power_law_1.2,0.058841598033905027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,power_law_1.2,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,power_law_1.2,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,power_law_1.2,0.08552320003509521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,power_law_1.2,0.11137919425964356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,power_law_1.2,0.13576960563659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,power_law_1.2,0.17987200021743774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.24528000354766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.34260480403900145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.5230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.7042943954467773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,power_law_1.01,0.9641599655151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.2192704200744628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.2759488105773926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.251142406463623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.3088512420654297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.4333439826965333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.3438336372375488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.4476032257080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.416096019744873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.4104576110839844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.5369983673095704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,power_law_1.01,1.511622428894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,power_law_1.2,0.23926401138305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,power_law_1.01,1.7083520889282227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,power_law_1.01,1.7026176452636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,power_law_1.01,1.8503488540649413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,power_law_1.01,1.8471488952636719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,power_law_1.01,2.098752021789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,power_law_1.01,2.407948875427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,power_law_1.01,2.8741056442260744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,power_law_1.01,3.13570556640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,power_law_1.01,4.260467147827148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,power_law_1.01,5.049516677856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,power_law_1.2,0.32940800189971925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,power_law_1.2,0.4009664058685303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,power_law_1.2,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,power_law_1.2,0.0695680022239685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,power_law_1.2,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,power_law_1.2,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,power_law_1.2,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,power_law_1.2,0.07775359749794006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,power_law_1.2,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,power_law_1.2,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,power_law_1.2,0.08398079872131348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,power_law_1.2,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,power_law_1.2,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,power_law_1.2,0.09164159893989562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,power_law_1.2,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,power_law_1.2,0.09900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,power_law_1.2,0.1072383999824524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,power_law_1.2,0.1272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.2,0.15212160348892212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.2,0.19062399864196777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.2,0.24499199390411378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.2,0.33729920387268064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.2,0.37861759662628175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.2,0.61113600730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.2,0.7579775810241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.2,1.1870976448059083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.2,1.5574848175048828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,power_law_1.2,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,power_law_1.2,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,power_law_1.2,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,power_law_1.2,0.12356480360031127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,power_law_1.2,0.1312064051628113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,power_law_1.2,0.13525760173797607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,power_law_1.2,0.13582719564437867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,power_law_1.2,0.1402176022529602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,power_law_1.2,0.14584319591522216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,power_law_1.2,0.1517184019088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,power_law_1.2,0.15623040199279786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,power_law_1.2,0.1646016001701355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,power_law_1.2,0.18891520500183107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.2,0.21448960304260253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.2,0.2601151943206787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.2,0.3161920070648193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.2,0.4181248188018799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.2,0.5414080142974853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.2,0.6928063869476319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.2,0.9742143630981446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.14614399671554565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.22047359943389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.3115839958190918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.3931328058242798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6067264080047607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.6318463802337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.6665215969085694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7072192192077636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.7173888206481933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.7473919868469239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.7651328086853028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8092608451843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.8073920249938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,power_law_1.2,0.8809215545654296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,power_law_1.2,0.9326144218444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,power_law_1.2,0.9630975723266602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.1184639930725098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.2384575843811034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.4698880195617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.575500774383545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,power_law_1.2,1.9767936706542968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.3617664337158204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.1577791213989257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,power_law_1.2,3.9811073303222657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.2,1.3085824012756349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.2,1.7089216232299804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,balanced,0.02553066611289978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,balanced,0.02369600037733714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,balanced,0.025562666356563568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,balanced,0.027450665831565857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,balanced,0.02792533238728841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,balanced,0.029472000896930695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,balanced,0.036042665441830955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,balanced,0.07021333277225494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,balanced,0.09521067142486572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,balanced,0.11849600076675415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,balanced,0.1481706698735555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,balanced,0.18440532684326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,balanced,0.04586666822433472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,balanced,0.048101335763931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,balanced,0.05089066425959269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,balanced,0.05605866511662801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,power_law_1.01,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,power_law_1.01,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,balanced,0.07241599758466084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,power_law_1.01,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,power_law_1.01,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,balanced,0.09058133761088054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,balanced,0.10652800401051839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,balanced,0.13847999771436056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,balanced,0.16768532991409302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,power_law_1.01,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,balanced,0.22724799315134683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,balanced,0.2823946674664815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,power_law_1.01,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,power_law_1.01,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,power_law_1.01,0.07223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,power_law_1.01,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,power_law_1.01,0.08407679796218873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,power_law_1.01,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,power_law_1.01,0.10813440084457397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,power_law_1.01,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,power_law_1.01,0.15132800340652466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,power_law_1.01,0.19325439929962157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,power_law_1.01,0.2493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,power_law_1.01,0.355129599571228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,power_law_1.01,0.44004478454589846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,balanced,0.07588266829649608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,balanced,0.10248000423113506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,balanced,0.15194132924079895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,balanced,0.14899733662605286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,balanced,0.15065067013104758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,balanced,0.15169066190719604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,balanced,0.15278933445612589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,balanced,0.15435199936230978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,balanced,0.15161599715550741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,balanced,0.15312533577283224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,balanced,0.15557333827018738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,balanced,0.15716266632080078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,balanced,0.15759999553362528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,balanced,0.16293866435686746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,balanced,0.1718719998995463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,balanced,0.17479467391967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,balanced,0.17972266674041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,balanced,0.2133386731147766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.23147199551264444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.27668799956639606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.3364853461583455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.44940801461537677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.5626826683680216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,balanced,0.7807466983795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.0146613121032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,balanced,1.4485920270284016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,balanced,1.980277379353841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,balanced,0.0842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,balanced,0.11643733580907185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,balanced,0.17715734243392944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,balanced,0.29470932483673096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,balanced,0.2933280070622762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,balanced,0.29872532685597736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,balanced,0.29872532685597736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,balanced,0.3001813292503357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,balanced,0.3019040028254191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,balanced,0.3044106761614482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,balanced,0.3064746658007304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,balanced,0.3095039923985799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,balanced,0.3184159994125366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,balanced,0.3186239997545878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,balanced,0.32684799035390216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,balanced,0.3394293387730916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.35195199648539227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.3828800121943156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.41518934567769367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.4724106788635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.5284479856491089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.6995573043823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,balanced,0.7994613647460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.1274080276489258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.3843572934468586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.03299199938774109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.14012160301208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.19355520009994506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.2869247913360596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,power_law_1.2,0.056460797786712646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,power_law_1.2,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,power_law_1.2,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,power_law_1.2,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,power_law_1.2,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,power_law_1.2,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,power_law_1.2,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.2,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.2,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.2,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.2,0.12663040161132813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.2,0.14448000192642213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.2,0.21016321182250977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.2,0.23446400165557862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.2,0.3530495882034302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.2,0.45240321159362795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.11615359783172607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.11905920505523682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.11907839775085449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.11416319608688355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.118995201587677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.11715840101242066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.13269120454788208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.1316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.16547199487686157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.2022144079208374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.21162240505218505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.26752638816833496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.3116224050521851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.4166848182678223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,power_law_1.2,0.47588481903076174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,power_law_1.2,0.6483456134796143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,power_law_1.2,0.8819583892822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.08355839848518372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.08583040237426758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.08725759983062745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.08355839848518372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.08178560137748718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.09943680167198181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.09989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.14188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.1805184006690979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.20502400398254395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.2888767957687378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.35394558906555174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,power_law_1.01,0.5023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,power_law_1.01,0.6847616195678711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.08849279880523682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.10591360330581664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.14586880207061767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.1542080044746399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.155731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.16700799465179444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.1737663984298706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.17596160173416137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.1813696026802063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.18843519687652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.19775999784469606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.20908160209655763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.23624320030212403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.2650496006011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.32522239685058596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.36337919235229493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.4650432109832764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.5596415996551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,0.7503615856170655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,0.939891242980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,1.325260829925537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,1.7221952438354493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,power_law_1.2,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,power_law_1.2,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,power_law_1.2,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,power_law_1.2,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,power_law_1.2,0.1180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,power_law_1.2,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,power_law_1.2,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,power_law_1.2,0.13137279748916625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,power_law_1.2,0.13816319704055785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,power_law_1.2,0.1447487950325012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,power_law_1.2,0.14956159591674806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,power_law_1.2,0.1579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,power_law_1.2,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,power_law_1.2,0.17944960594177245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,power_law_1.2,0.18350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,power_law_1.2,0.21430399417877197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,power_law_1.2,0.24238719940185546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,power_law_1.2,0.3006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,power_law_1.2,0.3507904052734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,power_law_1.2,0.46711039543151855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,power_law_1.2,0.5792384147644043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,power_law_1.2,0.7947519779205322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,power_law_1.2,1.018380832672119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,power_law_1.2,1.4480319976806642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,power_law_1.2,1.8851200103759767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,balanced,0.08303466439247131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,balanced,0.08592533071835835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,balanced,0.090421328941981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,balanced,0.09326400359471639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,balanced,0.0934933324654897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,balanced,0.09540266791979472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,balanced,0.10128532846768697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,balanced,0.10250666737556458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,balanced,0.10708266496658325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,balanced,0.12798399726549783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,balanced,0.12892267107963562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,balanced,0.16062933206558228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,balanced,0.16830400625864664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,balanced,0.23078932364781699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,balanced,0.2769920031229655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,balanced,0.38395198186238605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,balanced,0.48133333524068195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,balanced,0.6783520380655924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,balanced,0.8534453709920248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.03832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.05053439736366272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.08951680064201355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.10419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.13772799968719482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.1654144048690796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.22965760231018068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.2903167963027954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.4179520130157471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.04305280148983002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.55000319480896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.060198402404785155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.10514559745788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.1295040011405945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.15912959575653077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.18198399543762206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.06711040139198303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.06996480226516724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.09867519736289979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.1161344051361084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.1269376039505005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.15820800065994262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.1843840003013611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.23719680309295654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.2693056106567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.3836544036865234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.5265984058380127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.7619328022003173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.9132287979125977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.299398398399353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.3808255910873413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,power_law_1.2,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,power_law_1.2,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,power_law_1.2,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,power_law_1.2,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.2,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.2,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.2,0.11178239583969116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.2,0.13456000089645387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.2,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.2,0.2325824022293091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.2,0.3326143980026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.2,0.4085824012756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,balanced,0.056218668818473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,balanced,0.05866133173306783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,balanced,0.06057066718737284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,balanced,0.05959466596444448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,balanced,0.06374399860699971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,balanced,0.07005866865317027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,balanced,0.06836799780527751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,balanced,0.075354665517807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,balanced,0.08067200084527333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,balanced,0.10810133814811707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,balanced,0.13935466607411703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,balanced,0.1514400045077006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,balanced,0.21206400791803995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,balanced,0.2526879906654358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,balanced,0.3521920045216878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,balanced,0.4410933256149292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,balanced,0.0572266678015391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,balanced,0.07850666840871175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,balanced,0.07694933315118153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,balanced,0.07855999966462453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,balanced,0.07933866480986278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,balanced,0.0790826678276062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,balanced,0.07912000020345052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,balanced,0.08011200030644734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,balanced,0.0817440003156662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,balanced,0.09027199943860371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,balanced,0.09265599648157756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,balanced,0.09739733735720317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,balanced,0.10147733489672343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,balanced,0.10878933469454448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,balanced,0.12474133570988973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,balanced,0.13854933778444925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,balanced,0.16613333423932394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,balanced,0.19434134165445963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,balanced,0.2537920077641805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,balanced,0.31441599130630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,balanced,0.4249279896418254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,balanced,0.5410666863123575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.10351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.14616960287094116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.23826560974121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.21041278839111327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.2569535970687866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.25052800178527834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.2671999931335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.2768831968307495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.28474879264831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.27853438854217527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.29303679466247556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.2952192068099976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.31351680755615235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.29643518924713136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.30915839672088624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.3154239892959595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.3247936010360718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.3609663963317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.3689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.43438081741333007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.45475201606750487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.5200191974639893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,power_law_1.01,0.5850560188293457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,power_law_1.01,0.7203839778900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,power_law_1.01,0.8466560363769531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,power_law_1.01,1.0459712028503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,power_law_1.01,1.2858176231384277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,power_law_1.2,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,power_law_1.2,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,power_law_1.2,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,power_law_1.2,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,power_law_1.2,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,power_law_1.2,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,power_law_1.2,0.09086080193519593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,power_law_1.2,0.09870079755783082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,power_law_1.2,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.2,0.13776639699935914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.2,0.2007551908493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.2,0.2611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.2,0.3998975992202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.2,0.43615360260009767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.2,0.7619200229644776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.2,0.9320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.2,1.277248001098633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.2,1.830303955078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.030982398986816408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.0460671991109848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,power_law_1.2,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,power_law_1.2,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,power_law_1.2,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,power_law_1.2,0.03681919872760773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,power_law_1.2,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,power_law_1.2,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,power_law_1.2,0.11364480257034301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,power_law_1.2,0.14869760274887084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.22359681129455566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,power_law_1.2,0.21502079963684081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,power_law_1.2,0.25036160945892333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.2839871883392334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,power_law_1.2,0.4392255783081055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.40076799392700196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,power_law_1.2,0.5116991996765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.515225601196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.06384639739990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.11916160583496094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.12121599912643433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.12625279426574706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.12735999822616578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.129312002658844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.13144320249557495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.14384000301361083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.16036479473114013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.1579327940940857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.16227840185165404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.21742720603942872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.21132159233093262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.27606399059295655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.3388351917266846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,power_law_1.2,0.43258237838745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,power_law_1.2,0.554099178314209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,power_law_1.2,0.8155776023864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,power_law_1.2,1.003654384613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,power_law_1.2,1.44966402053833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,power_law_1.2,1.863692855834961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,balanced,0.03378133227427801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,balanced,0.040576001008351646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,balanced,0.059690664211908974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,balanced,0.08582400282224019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,balanced,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,balanced,0.08912000060081482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,balanced,0.08971200386683147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,balanced,0.09142399827639262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,balanced,0.09523733456929524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,balanced,0.09409600496292114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,balanced,0.09909333785374959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,balanced,0.1048426628112793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,balanced,0.1113920013109843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,balanced,0.11619200309117635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,balanced,0.11687466502189636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,balanced,0.13238933682441711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,balanced,0.1471733351548513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,balanced,0.18386133511861166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,balanced,0.19644800821940103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,balanced,0.278330663839976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,balanced,0.32128532727559406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,balanced,0.4655253489812215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,balanced,0.5912479956944784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,balanced,0.8502026398976644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,balanced,1.112277348836263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.09443839788436889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.10091520547866821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.10943360328674316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.10380799770355224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.10638079643249512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.11213439702987671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.11759359836578369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.13096319437026976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.1349951982498169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.14202239513397216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.1561344027519226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.17905919551849364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.21025919914245605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.2359679937362671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.30556159019470214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.35329279899597166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.49015040397644044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.6133312225341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.8155776023864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.0402751922607423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.12042880058288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.12317440509796143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.12714240550994874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.13137279748916625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.13827199935913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.14654719829559326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.15312000513076782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.16595200300216675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.19835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.22738559246063234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.28832640647888186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.347110390663147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.4771455764770508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.597049617767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,0.8485952377319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.0880767822265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,1.567628765106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,2.0517568588256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.0671231985092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.06911360025405884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.0720575988292694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.09365760087966919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.09884799718856811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.11306240558624267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.12072960138320923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.14627840518951415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.16420480012893676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.24483840465545653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.2587199926376343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.36563839912414553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.49649920463562014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,power_law_1.2,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,power_law_1.2,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,power_law_1.2,0.058815997838973996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,power_law_1.2,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,power_law_1.2,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,power_law_1.2,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,power_law_1.2,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,power_law_1.2,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,power_law_1.2,0.08851199746131896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,power_law_1.2,0.09210879802703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,power_law_1.2,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,power_law_1.2,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.6894144058227539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,power_law_1.2,0.10405759811401367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,power_law_1.2,0.10385279655456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,power_law_1.2,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,power_law_1.2,0.11571840047836304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,power_law_1.2,0.12422399520874024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,power_law_1.2,0.14247679710388184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,power_law_1.2,0.15762560367584227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,power_law_1.2,0.199072003364563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,power_law_1.2,0.23420159816741942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,power_law_1.2,0.310809588432312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,power_law_1.2,0.3966655969619751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,power_law_1.2,0.5467199802398681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,power_law_1.2,0.7168255805969238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,power_law_1.2,0.9467455863952636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.07247999906539918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.09178879857063293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.09663360118865967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.10695040225982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.10970879793167114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.11404800415039062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.10972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.1260800004005432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.13720959424972534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.1508031964302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.15940480232238768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.18870400190353392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.2105855941772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.26481919288635253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.31629440784454343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.4413760185241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.5581567764282227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.7620927810668945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,0.9261183738708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.9193408012390136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,power_law_1.2,1.312345600128174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,balanced,0.0683840016523997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,balanced,0.09409600496292114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,balanced,0.1134986678759257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,balanced,0.12664000193277994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,balanced,0.12814933061599731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,balanced,0.12988266348838806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,balanced,0.1276853382587433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,balanced,0.12922133008639017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,balanced,0.12915733456611633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,balanced,0.12985066572825113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,balanced,0.13429333766301474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,balanced,0.13356266419092813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,balanced,0.1390666663646698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,balanced,0.14355199535687765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,balanced,0.1460533340771993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,balanced,0.1588159998257955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,balanced,0.17075733343760172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,balanced,0.2118133306503296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,balanced,0.22261333465576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,balanced,0.3051626682281494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,balanced,0.3035200039545695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,balanced,0.4198400179545085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,balanced,0.4673279921213786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,balanced,0.6596320072809855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,balanced,0.7809120019276937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,power_law_1.2,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,power_law_1.2,0.11623680591583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,power_law_1.2,0.14699519872665406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,power_law_1.2,0.1618880033493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,power_law_1.2,0.17611520290374755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,power_law_1.2,0.19633280038833617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,power_law_1.2,0.24456319808959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,power_law_1.2,0.2430527925491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,power_law_1.2,0.24709761142730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,power_law_1.2,0.254694390296936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,power_law_1.2,0.2631103992462158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,power_law_1.2,0.27015678882598876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,power_law_1.2,0.2707520008087158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,power_law_1.2,0.28332159519195554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,power_law_1.2,0.2835776090621948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,power_law_1.2,0.28682880401611327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,power_law_1.2,0.304582405090332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,power_law_1.2,0.3363071918487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.2,0.3724992036819458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.2,0.4394559860229492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.2,0.5186944007873535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.2,0.6302271842956543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.2,0.7797952175140381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.2,1.0426688194274902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.2,1.3188480377197265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.2,1.8927104949951172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.2,2.458086395263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,power_law_1.01,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,power_law_1.01,0.1428287982940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,power_law_1.01,0.17690240144729613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,power_law_1.01,0.21123840808868408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,power_law_1.01,0.21068799495697021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,power_law_1.01,0.21490559577941895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,power_law_1.01,0.2198848009109497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,power_law_1.01,0.2229248046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,power_law_1.01,0.22391679286956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,power_law_1.01,0.2325824022293091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,power_law_1.01,0.24102399349212647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,power_law_1.01,0.24737279415130614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,power_law_1.01,0.25943679809570314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,power_law_1.01,0.2730304002761841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,power_law_1.01,0.3149375915527344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,power_law_1.01,0.33649919033050535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,power_law_1.01,0.3158911943435669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,power_law_1.01,0.3722239971160889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,power_law_1.01,0.4476287841796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,power_law_1.01,0.5885568141937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,power_law_1.01,0.6989247798919678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,power_law_1.01,0.9415360450744629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,power_law_1.01,1.124448013305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,power_law_1.01,1.4819199562072753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,power_law_1.01,2.1745344161987306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,power_law_1.01,2.7051519393920898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,power_law_1.01,3.8582977294921874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.13312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.1827839970588684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.20003199577331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.3168832063674927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.4389632225036621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,balanced,0.07333333293596904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,balanced,0.0981813371181488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,balanced,0.14695466558138529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,balanced,0.2453546722730001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,balanced,0.4395253260930379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,balanced,0.44164268175760907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,balanced,0.44333867232004803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,balanced,0.4435253143310547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,balanced,0.4458880027135213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,balanced,0.4476319948832194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,balanced,0.45394666989644367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,balanced,0.45713599522908527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,balanced,0.4594293435414632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,balanced,0.4681973457336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,balanced,0.4716586669286092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,balanced,0.4784746567408244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,balanced,0.49933334191640216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,balanced,0.5134933392206827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,balanced,0.5490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,balanced,0.5890026489893595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,balanced,0.6562933524449667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,balanced,0.725717306137085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,balanced,0.901151974995931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,balanced,0.9814773400624593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,balanced,1.432026704152425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,balanced,1.6217919985453289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.07117440104484558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.1371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.16480000019073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.20419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.2563071966171265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.35685760974884034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,balanced,0.06992533306280772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,balanced,0.09642133116722107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,balanced,0.14632532993952432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,balanced,0.14733333388964334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,balanced,0.14721600214640299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,balanced,0.14871999621391296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,balanced,0.14779200156529745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,balanced,0.14944533507029215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,balanced,0.15010133385658264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,balanced,0.15212800105412802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,balanced,0.15100266536076865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,balanced,0.1574560006459554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,balanced,0.15854400396347046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,balanced,0.16057599584261575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,balanced,0.16663466890652975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,balanced,0.17126933733622232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,balanced,0.17933332920074463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,balanced,0.19358932971954346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,balanced,0.21355734268824259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,balanced,0.2296746571858724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,balanced,0.28545600175857544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,balanced,0.3218933343887329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,balanced,0.43224533398946124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,balanced,0.497162659962972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.47307519912719725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,balanced,0.06311466793219249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,balanced,0.08993599812189738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,balanced,0.11183466513951619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,balanced,0.11356266339619954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,balanced,0.1163040002187093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,balanced,0.11433066924413045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,balanced,0.11717333396275838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,balanced,0.11618133385976155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,balanced,0.11682666341463725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,balanced,0.11958932876586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,balanced,0.1199626624584198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,balanced,0.1280639966328939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,balanced,0.12984533111254373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,balanced,0.13317867120107016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,balanced,0.13809067010879517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,balanced,0.14486933747927347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,balanced,0.1604639987150828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,balanced,0.17081065972646078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,balanced,0.20409599939982095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,balanced,0.2264639933904012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,balanced,0.29572800795237225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,balanced,0.3442933162053426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,balanced,0.47898133595784503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,balanced,0.6129226684570312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.6917695999145508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,0.9215295791625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,power_law_1.01,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,power_law_1.01,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,power_law_1.01,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,power_law_1.01,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,power_law_1.01,0.07153279781341552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,power_law_1.01,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,power_law_1.01,0.0807424008846283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,power_law_1.01,0.09021440148353577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,power_law_1.01,0.11653120517730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,power_law_1.01,0.1355839967727661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,power_law_1.01,0.16660480499267577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,power_law_1.01,0.19698560237884521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,power_law_1.01,0.2601599931716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,power_law_1.01,0.35228159427642824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,power_law_1.01,0.4877120018005371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,power_law_1.01,0.6340032100677491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,power_law_1.01,0.9154623985290528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,power_law_1.01,1.1934335708618165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,balanced,0.04808000226815542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,balanced,0.06086933116118113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,balanced,0.06562133133411407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,balanced,0.06538133323192596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,balanced,0.06700266897678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,balanced,0.06669866542021434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,balanced,0.06731200218200684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,balanced,0.06699199974536896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,balanced,0.06809600194295247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,balanced,0.07008000214894612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,balanced,0.07115200161933899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,balanced,0.0718399981657664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,balanced,0.08364799618721008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,balanced,0.08834667007128398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,balanced,0.09872532884279887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,balanced,0.10564800103505452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,balanced,0.12786666552225748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,balanced,0.1518826683362325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,balanced,0.20283732811609903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,balanced,0.24871466557184854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,balanced,0.34432534376780194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,balanced,0.4469546476999919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,balanced,0.6563413143157959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,balanced,0.8541119893391927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,balanced,0.05930666625499725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,balanced,0.08090666433175404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,balanced,0.1097866694132487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,balanced,0.15845866998036703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,balanced,0.1606826682885488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,balanced,0.16378666957219443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,balanced,0.16306133071581522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,balanced,0.16394666830698648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,balanced,0.16781866550445557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,balanced,0.16952532529830933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,balanced,0.1702400048573812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,balanced,0.1743839979171753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,balanced,0.18539732694625854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,balanced,0.1845759948094686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,balanced,0.18636800845464072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,balanced,0.20407466093699136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,balanced,0.21737066904703775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,balanced,0.2502453327178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,balanced,0.27983999252319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,balanced,0.3418346643447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,balanced,0.4044959942499797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,balanced,0.5606773296991984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,balanced,0.6813279787699381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,balanced,0.9774933656056722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,balanced,1.246010700861613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.02868480086326599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.026399999856948853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.02815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.02881920039653778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.029708799719810487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.1375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,power_law_1.2,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,power_law_1.2,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,power_law_1.2,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,power_law_1.2,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,power_law_1.2,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,power_law_1.2,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.21962239742279052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,power_law_1.2,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,power_law_1.2,0.14933120012283324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,power_law_1.2,0.1897279977798462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,power_law_1.2,0.25725440979003905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.3025536060333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,power_law_1.2,0.3596224069595337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,power_law_1.2,0.45840001106262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,balanced,0.07743999858697255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,balanced,0.10548266768455505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,balanced,0.15755200386047363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,balanced,0.264138658841451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,balanced,0.31753067175547284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,balanced,0.3227360049883525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,balanced,0.3206613262494405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,balanced,0.3234773278236389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,balanced,0.32630399862925213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,balanced,0.3296479980150859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,balanced,0.33311466375986737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,balanced,0.3394506772359212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,balanced,0.35066131750742596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,balanced,0.3603733380635579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,balanced,0.3694399992624919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,balanced,0.37746667861938477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,balanced,0.4121439854303996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,balanced,0.43649065494537354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,balanced,0.5412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,balanced,0.5682080189387003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,balanced,0.8242666721343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,balanced,0.8557013670603434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,balanced,1.3819519678751628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,balanced,1.4507999420166016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,balanced,2.3169172604878745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,balanced,2.707834561665853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,power_law_1.2,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,power_law_1.2,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,power_law_1.2,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,power_law_1.2,0.044377601146698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,power_law_1.2,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,power_law_1.2,0.06551679968833923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,power_law_1.2,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,power_law_1.2,0.11285760402679443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,power_law_1.2,0.14357759952545165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,power_law_1.2,0.18067840337753296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,power_law_1.2,0.24766080379486083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,power_law_1.2,0.3335103988647461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,power_law_1.2,0.4588287830352783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,power_law_1.2,0.6160128116607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.055404800176620486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.06353920102119445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.06714239716529846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.1030784010887146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.12253439426422119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.15167360305786132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.18106240034103394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.2331712007522583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.278003191947937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.41498241424560545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.5186240196228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,power_law_1.01,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,power_law_1.01,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,power_law_1.01,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,power_law_1.01,0.0633408010005951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,power_law_1.01,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,power_law_1.01,0.09040639996528625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,power_law_1.01,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,power_law_1.01,0.12337919473648071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,power_law_1.01,0.1345023989677429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,power_law_1.01,0.2048192024230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,power_law_1.01,0.25992960929870607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,power_law_1.01,0.3632447957992554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,power_law_1.01,0.45754241943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.1154047966003418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.1756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.21418240070343017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.2070591926574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.21409919261932372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.21907200813293456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.2199552059173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.23304319381713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.2211519956588745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.2482304096221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.23808000087738038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.25778560638427733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.2785279989242554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.29356160163879397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.3153408050537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.3463295936584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,power_law_1.01,0.3957567930221558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,power_law_1.01,0.48967680931091306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,power_law_1.01,0.6294528007507324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,power_law_1.01,0.7776127815246582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,power_law_1.01,0.8857215881347656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,power_law_1.01,1.6770559310913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,power_law_1.01,2.004761505126953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.07140480279922486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.15966720581054689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.1673792004585266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.19731839895248413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.2075455904006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.21094400882720948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.20824320316314698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.21276159286499025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.22517759799957277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.2302016019821167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.23992319107055665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.25402240753173827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.25481600761413575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.2766848087310791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.30227839946746826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.36097280979156493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,power_law_1.01,2.992870330810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.41001601219177247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.5072192192077637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.6334847927093505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,0.8729215621948242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.1084416389465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.514515209197998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,1.9229888916015625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,power_law_1.01,3.5772480010986327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08318719863891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.15301120281219482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.21239039897918702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.27723519802093505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.33534080982208253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.449510383605957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.473356819152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.48862080574035643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5039999961853028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5349696159362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.5479040145874023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.5680768013000488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.5881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.611359977722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.631059217453003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.6986559867858887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.7546688079833984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,power_law_1.2,0.8835583686828613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,power_law_1.2,0.9706111907958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.1715456008911134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.4564607620239258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,power_law_1.2,1.8371200561523438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.0959232330322264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,power_law_1.2,2.7578752517700194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,power_law_1.2,3.5351936340332033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,power_law_1.2,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,power_law_1.2,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,power_law_1.2,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,power_law_1.2,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,power_law_1.2,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,power_law_1.2,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,power_law_1.2,0.06581119894981384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,power_law_1.2,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,power_law_1.2,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,power_law_1.2,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,power_law_1.2,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.2,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.2,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.2,0.13601919412612914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.2,0.1676543951034546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.2,0.19466880559921265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.2,0.291596794128418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.2,0.34882559776306155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.2,0.4906816005706787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.2,0.6196095943450928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.058931201696395874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.09872000217437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.1068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.12479360103607177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.12891520261764527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.12288639545440674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.12275199890136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.1295040011405945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.13263360261917115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.15690239667892455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.163372802734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.1592319965362549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.18028160333633422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.1953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.23627519607543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.2609215974807739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.3477504014968872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.4085184097290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.5797632217407227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.6995456218719482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,0.9562175750732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.3393983840942383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,power_law_1.01,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,power_law_1.01,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,power_law_1.01,0.030188798904418945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,power_law_1.01,0.028972798585891725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,power_law_1.01,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,power_law_1.01,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,power_law_1.01,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,power_law_1.01,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,power_law_1.01,0.14338560104370118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,power_law_1.01,0.15447039604187013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,power_law_1.01,0.21982080936431886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,power_law_1.01,0.28770558834075927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,balanced,0.0339626669883728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,balanced,0.04098666707674662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,balanced,0.06588266789913177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,balanced,0.10133333007494609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,balanced,0.13319466511408487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,balanced,0.16593066851298013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,balanced,0.22196267048517862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,balanced,0.27356799443562824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,balanced,0.036703998843828835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,balanced,0.055231998364130654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,balanced,0.07766399780909221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,balanced,0.11970133582750957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,balanced,0.20285866657892862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,balanced,0.2036906679471334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,balanced,0.2063360015551249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,balanced,0.20545599857966104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,balanced,0.20487467447916666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,balanced,0.2067306637763977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,balanced,0.20784000555674234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,balanced,0.21113600333531699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,balanced,0.21306133270263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,balanced,0.2161653240521749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,balanced,0.22080532709757486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,balanced,0.2249280015627543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,balanced,0.23162132501602173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,balanced,0.2436479926109314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,balanced,0.2552000085512797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,balanced,0.2959199945131938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,balanced,0.3131733338038127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,balanced,0.3858400185902913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,balanced,0.4224693377812703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,balanced,0.5627679824829102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,balanced,0.6594986518224081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,balanced,0.9199413458506266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,balanced,1.1749760309855144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,balanced,0.05199466645717621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,balanced,0.07778133451938629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,balanced,0.10149866342544556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,balanced,0.10357333223025005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,balanced,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,balanced,0.10402133067448933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,balanced,0.10648000240325928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,balanced,0.107205331325531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,balanced,0.10876267155011494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,balanced,0.10910933216412862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,balanced,0.1170186698436737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,balanced,0.11728533109029134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,balanced,0.12542399764060974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,balanced,0.13482133547465006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.14562666416168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.1685439944267273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.18170666694641113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.21967466672261557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.2602399984995524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.343231995900472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.4532266855239868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.6136693159739176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.793999989827474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,power_law_1.01,0.07727360129356384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,power_law_1.01,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,power_law_1.01,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,power_law_1.01,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,power_law_1.01,0.10913280248641968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,power_law_1.01,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,power_law_1.01,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,power_law_1.01,0.11192959547042847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,power_law_1.01,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,power_law_1.01,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,power_law_1.01,0.11920640468597413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,power_law_1.01,0.12115199565887451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,power_law_1.01,0.12444159984588624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,power_law_1.01,0.13634560108184815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.01,0.14707839488983154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.01,0.173363196849823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.01,0.19493119716644286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.01,0.2489151954650879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.01,0.3100287914276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.01,0.433241605758667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.01,0.5255424022674561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.01,0.7870528221130371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.01,1.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,balanced,0.07495999832948048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,balanced,0.09845333298047383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,balanced,0.1497173309326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,balanced,0.24342934290568033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,balanced,0.4445493221282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,balanced,0.8293866316477457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,balanced,1.2202506860097249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,balanced,1.2243626912434895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,balanced,1.2270879745483398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,balanced,1.2293866475423176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,balanced,1.2317279974619548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,balanced,1.2450986703236897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,balanced,1.2502293586730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,balanced,1.2621386845906575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,balanced,1.2749706904093425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,balanced,1.2870826721191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,balanced,1.3085652987162273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,balanced,1.35150941212972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,balanced,1.385599931081136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,balanced,1.468666712443034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,balanced,1.5400427182515461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,balanced,1.6741013526916504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,balanced,1.8115040461222331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,power_law_1.01,0.028992000222206115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,balanced,2.112581411997477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,power_law_1.01,0.0281792014837265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,balanced,2.559173266092936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,power_law_1.01,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,power_law_1.01,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,balanced,3.1254027684529624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,balanced,4.192917188008626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,power_law_1.01,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,power_law_1.01,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,power_law_1.01,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,power_law_1.01,0.040479999780654904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,power_law_1.01,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,power_law_1.01,0.058841598033905027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,power_law_1.01,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,power_law_1.01,0.11163519620895386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,power_law_1.01,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,power_law_1.01,0.206278395652771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,power_law_1.01,0.27509119510650637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.085343998670578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.1147007942199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.13115520477294923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.19587839841842652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.17014399766921998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.2322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.21507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,power_law_1.2,0.318668794631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.32928640842437745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,power_law_1.2,0.40677762031555176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,power_law_1.2,0.4133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,balanced,0.04911466439565023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,balanced,0.08224533498287201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,balanced,0.11782399813334148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,balanced,0.11751466989517212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,balanced,0.1239520013332367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,balanced,0.12310933073361714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,balanced,0.12237333257993062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,balanced,0.12688000003496805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,balanced,0.1255519986152649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,balanced,0.13111467162768045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,balanced,0.13365333278973898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,balanced,0.1356053352355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,balanced,0.14356266458829245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,balanced,0.14749866724014282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,balanced,0.156549334526062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,balanced,0.17828265825907388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,balanced,0.1993066668510437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,balanced,0.2423093318939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,balanced,0.28296534220377606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,balanced,0.3961493174235026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,balanced,0.47992531458536786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,balanced,0.6923253536224365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,balanced,0.8738719622294108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,balanced,1.2817973295847576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,balanced,1.6809760729471843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.24483199119567872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.40295681953430174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.5608128070831299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.6239935874938964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.8328448295593261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.8649408340454101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.8911999702453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,power_law_1.01,0.9238783836364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,power_law_1.01,0.9722047805786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,power_law_1.01,0.9943615913391113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.0108927726745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.0344191551208497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.0790847778320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.1216256141662597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.11612160205841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.1560960054397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.2011072158813476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.19624320268630982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.25870718955993655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.2877887725830077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.34530560970306395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.47557759284973145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.4977791786193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,power_law_1.01,0.5620031833648682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.4629311561584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,power_law_1.01,1.6792320251464843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.0000192642211916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,power_law_1.01,2.5022016525268556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,power_law_1.01,2.7082752227783202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,power_law_1.01,3.4706817626953126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,power_law_1.01,4.282035064697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,power_law_1.2,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,power_law_1.2,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,power_law_1.2,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,power_law_1.2,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,power_law_1.2,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,power_law_1.2,0.1360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,power_law_1.2,0.17111040353775026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,power_law_1.2,0.269484806060791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,power_law_1.2,0.32401280403137206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.12091519832611083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.14357759952545165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.1715456008911133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.21059200763702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.2131903886795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.2173759937286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.22351360321044922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.23082239627838136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.23764479160308838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.24389119148254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.25080959796905516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.2636159896850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.26818559169769285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.28815360069274903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.3333695888519287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.3715008020401001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.45890560150146487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5442111968994141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.7160128116607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.8782719612121582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.222764778137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.5649984359741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.252620887756348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,2.9344192504882813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,power_law_1.2,0.42501120567321776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,power_law_1.2,0.5885248184204102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.1407807946205139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.17799680233001708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.2320319890975952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,power_law_1.2,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,power_law_1.2,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,power_law_1.2,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,power_law_1.2,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,power_law_1.2,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,power_law_1.2,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,power_law_1.2,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,power_law_1.2,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,power_law_1.2,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,power_law_1.2,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,power_law_1.2,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,power_law_1.2,0.14968960285186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,power_law_1.2,0.19322240352630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,power_law_1.2,0.23386878967285157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,power_law_1.2,0.3563647985458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,power_law_1.2,0.4957888126373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,power_law_1.2,0.6780543804168702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,power_law_1.01,0.05886080265045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,power_law_1.01,0.06881920099258423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,power_law_1.01,0.0781503975391388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,power_law_1.01,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,power_law_1.01,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,power_law_1.01,0.10985599756240845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,power_law_1.01,0.1327679991722107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,power_law_1.01,0.16852480173110962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,power_law_1.01,0.21505279541015626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,power_law_1.01,0.24357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,power_law_1.01,0.36703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,power_law_1.01,0.4423488140106201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,power_law_1.01,0.6511295795440674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,power_law_1.01,0.8294079780578614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,1,balanced,0.1597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,1,balanced,0.24132800102233887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,1,balanced,0.43303998311360675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,1,balanced,0.43729599316914874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,1,balanced,0.4375466505686442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,1,balanced,0.43905067443847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,1,balanced,0.43851200739542645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,1,balanced,0.43884801864624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,1,balanced,0.44149335225423175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,1,balanced,0.4403146505355835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,1,balanced,0.4418293237686157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,1,balanced,0.44628798961639404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,1,balanced,0.44812266031901044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,1,balanced,0.45278934637705487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,1,balanced,0.46300800641377765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,1,balanced,0.4665120045344035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,1,balanced,0.47651731967926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,1,balanced,0.6041546662648519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,1,balanced,0.6271786689758301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,1,balanced,0.8601706822713217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,1,balanced,1.0753013292948406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,1,balanced,1.537109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,1,balanced,2.0613226890563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,1,balanced,2.9705705642700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,1,balanced,3.881349245707194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,1,balanced,5.895061492919922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,1,balanced,7.715264002482097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,balanced,0.04133866727352142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,balanced,0.043749332427978516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,balanced,0.05083199838797251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,balanced,0.05089599887530009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,balanced,0.05806933343410492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,balanced,0.06797866523265839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,balanced,0.07833066582679749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,balanced,0.08441600203514099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,balanced,0.10317867000897725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,balanced,0.11129066348075867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,balanced,0.13828266660372415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,balanced,0.16856000820795694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,balanced,0.239519993464152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,balanced,0.2994239926338196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,balanced,0.44363200664520264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,balanced,0.5739039977391561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,balanced,0.05867200096448263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,balanced,0.1332586705684662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,balanced,0.13456533352533975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,balanced,0.13849600156148276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,balanced,0.13915200034777322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,balanced,0.141567995150884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,balanced,0.15038933356602988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,balanced,0.1471999982992808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,balanced,0.15481600165367126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,balanced,0.1551413337389628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,balanced,0.16210666298866272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,balanced,0.1706506609916687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,balanced,0.17799999316533408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,balanced,0.18710400660832724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,balanced,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,balanced,0.23997332652409872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,balanced,0.30833067496617633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,balanced,0.34222400188446045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,balanced,0.47882131735483807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,balanced,0.5723466475804647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,balanced,0.8140693505605062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,balanced,1.0391573111216228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,balanced,1.5044213930765789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,balanced,1.9568212827046711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,balanced,0.04238399863243103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,balanced,0.08618133266766866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,balanced,0.08302933474381764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,balanced,0.08551999926567078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,balanced,0.08771733442942302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,balanced,0.0922826627890269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,balanced,0.09469333291053772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,balanced,0.10000000397364299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,balanced,0.1042080024878184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,balanced,0.11209600170453389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,balanced,0.11408533652623494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,balanced,0.12556800246238708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,balanced,0.1482186714808146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,balanced,0.16759467124938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,balanced,0.23137599229812622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,balanced,0.2633066574732463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,balanced,0.3863946596781413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,balanced,0.47908798853556317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,balanced,0.7030666669209799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,balanced,0.9140480359395345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,balanced,1.3530774116516113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,balanced,1.7582987149556477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,balanced,0.07057600220044453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,balanced,0.09252267082532246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,balanced,0.12186132868131001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,balanced,0.18263467152913412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,balanced,0.3026346762975057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,balanced,0.5422826608022054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,balanced,0.5407093365987142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,balanced,0.5465386708577474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,balanced,0.5449920097986857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5494399865468343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5505813360214233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5561013221740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5586506525675455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5694400072097778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5757973194122314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5772533416748047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5879733165105184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,balanced,0.6096906661987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.633792002995809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,balanced,0.6780532995859782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.7392853101094564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,balanced,0.8304746945699056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,balanced,0.917797327041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.2449226379394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.3790507316589355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,balanced,1.9816959698994954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.448901335398356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.058841598033905027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.05934079885482788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.06443520188331604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.10911999940872193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.13783040046691894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.16789120435714722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.21994240283966066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.25530879497528075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.3617408037185669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.46657280921936034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,power_law_1.01,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,power_law_1.01,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,power_law_1.01,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,power_law_1.01,0.06723840236663818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,power_law_1.01,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,power_law_1.01,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,power_law_1.01,0.09005439877510071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,power_law_1.01,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,power_law_1.01,0.09267839789390564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,power_law_1.01,0.10589439868927002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.01,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.01,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.01,0.19479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.01,0.229420804977417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.01,0.3247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.01,0.4193408012390137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.01,0.6133887767791748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.01,0.8267840385437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,balanced,0.061008001367251076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,balanced,0.07810133198897044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,balanced,0.10071999828020732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,balanced,0.13384000460306802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,balanced,0.13479999701182047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,balanced,0.13780267039934793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,balanced,0.1355893313884735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,balanced,0.13779200116793314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,balanced,0.13784533739089966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,balanced,0.13806933164596558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,balanced,0.13919466733932495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,balanced,0.1393119990825653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,balanced,0.1439520021279653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,balanced,0.1464853286743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,balanced,0.14781866470972696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,balanced,0.15213333566983542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,balanced,0.15987199544906616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,balanced,0.175653338432312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,balanced,0.18818666537602743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,balanced,0.2178186575571696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,balanced,0.24438933531443277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,balanced,0.32843732833862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,balanced,0.3882346550623576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,balanced,0.548245350519816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,balanced,0.7144532998402914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.09434239864349366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.09944959878921508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.12540800571441652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.14703999757766723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.1786687970161438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.24035840034484862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.28878719806671144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.39559040069580076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.5472576141357421
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.7437695980072021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.9652928352355957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,power_law_1.2,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,power_law_1.2,0.08780159950256347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,power_law_1.2,0.08958079814910888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,power_law_1.2,0.09618560075759888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,power_law_1.2,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,power_law_1.2,0.1035647988319397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,power_law_1.2,0.10299520492553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,power_law_1.2,0.11990400552749633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,power_law_1.2,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,power_law_1.2,0.14260480403900147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,power_law_1.2,0.14968960285186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,power_law_1.2,0.17114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,power_law_1.2,0.2266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,power_law_1.2,0.27910399436950684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,power_law_1.2,0.33464961051940917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,power_law_1.2,0.4739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,power_law_1.2,0.5417600154876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,power_law_1.2,0.7847551822662353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,power_law_1.2,0.8928128242492676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,power_law_1.2,1.4709376335144042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,power_law_1.2,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,power_law_1.2,0.0286655992269516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,power_law_1.2,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,power_law_1.2,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,power_law_1.2,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,power_law_1.2,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.2,0.0781503975391388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.2,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.2,0.13270399570465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.2,0.19217920303344727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.2,0.2502784013748169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.2,0.34155519008636476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.2,0.5065216064453125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,power_law_1.01,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,power_law_1.01,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,power_law_1.01,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,power_law_1.01,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,power_law_1.01,0.06378239989280701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,power_law_1.01,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,power_law_1.01,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,power_law_1.01,0.09520639777183533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,power_law_1.01,0.10551680326461792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,power_law_1.01,0.12825599908828736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,power_law_1.01,0.1478271961212158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,power_law_1.01,0.18672640323638917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,power_law_1.01,0.23466880321502687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,power_law_1.01,0.3150079965591431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,power_law_1.01,0.4192063808441162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,power_law_1.01,0.6023680210113526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,power_law_1.01,0.7226560115814209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,power_law_1.2,2.378835105895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.2,0.8024831771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.2,1.1297280311584472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,balanced,0.05130666494369507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,balanced,0.07271466652552287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,balanced,0.09549333651860555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,balanced,0.13222400347391763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,balanced,0.15596266587575278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,balanced,0.1569386621316274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,balanced,0.15891733765602112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,balanced,0.15820266803105673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,balanced,0.16060266892115274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,balanced,0.15913066267967224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,balanced,0.15928533673286438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,balanced,0.16274666786193848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,balanced,0.1633333365122477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,balanced,0.17067732413609824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,balanced,0.17113065719604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,balanced,0.17761067549387613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,balanced,0.19954133033752441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,balanced,0.20358399550120035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,balanced,0.2756693363189697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,balanced,0.2605173389116923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,balanced,0.3616906801859538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,balanced,0.34225066502888996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,balanced,0.47629865010579425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,balanced,0.5335466861724854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,balanced,0.765775998433431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,balanced,0.8874346415201823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,balanced,0.04012800008058548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,balanced,0.041221333046754204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,balanced,0.04714133342107137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,balanced,0.04961599906285604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,balanced,0.05709866682688395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,balanced,0.07646933197975159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,balanced,0.08311999837557475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,balanced,0.10654933253924052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,balanced,0.12390933434168498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.09079679846763611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.11162240505218506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.17861119508743287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.22043519020080565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.3468096017837524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.3657151937484741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,balanced,0.05357333521048228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,balanced,0.1011306643486023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,balanced,0.09678399562835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,balanced,0.09921066959698994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,balanced,0.0993333359559377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,balanced,0.0993386705716451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,balanced,0.10371733705202739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,balanced,0.10515200098355611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,balanced,0.10751466949780782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,balanced,0.11508267124493916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,balanced,0.11889066298802693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,balanced,0.12753066420555115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,balanced,0.14512000481287637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,balanced,0.16193600495656332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,balanced,0.1893600026766459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,balanced,0.22358399629592896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,balanced,0.2807786663373311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,balanced,0.35925865173339844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,balanced,0.4954293171564738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,balanced,0.6615413427352905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,balanced,0.9461920261383057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,balanced,1.2110133171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.04051199853420258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,power_law_1.2,0.07863680124282837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,power_law_1.2,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,power_law_1.2,0.09088000059127807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,power_law_1.2,0.10611840486526489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,power_law_1.2,0.1128767967224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,power_law_1.2,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,power_law_1.2,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,power_law_1.2,0.11704319715499878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.11790720224380494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,power_law_1.2,0.11900160312652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,power_law_1.2,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,power_law_1.2,0.12352639436721802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,power_law_1.2,0.13708159923553467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,power_law_1.2,0.14787839651107787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,power_law_1.2,0.15330560207366944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,power_law_1.2,0.18165760040283202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,power_law_1.2,0.19809919595718384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,power_law_1.2,0.2310528039932251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,power_law_1.2,0.26869759559631345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,power_law_1.2,0.3463871955871582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,power_law_1.2,0.4394495964050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,power_law_1.2,0.576639986038208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,power_law_1.2,0.7118847846984864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.15217920541763305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.18495999574661254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,power_law_1.2,1.0958271980285645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,power_law_1.2,1.4889023780822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.2529344081878662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.28488318920135497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.4820223808288574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.5370751857757569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,power_law_1.2,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,power_law_1.2,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,power_law_1.2,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,power_law_1.2,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,power_law_1.2,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,power_law_1.2,0.05000320076942444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,power_law_1.2,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,power_law_1.2,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,power_law_1.2,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,power_law_1.2,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,power_law_1.2,0.09724159836769104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,power_law_1.2,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,power_law_1.2,0.17017600536346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,power_law_1.2,0.23393919467926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,power_law_1.2,0.28970239162445066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,power_law_1.2,0.4261824131011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,power_law_1.2,0.5670911788940429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.12577279806137084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.15391360521316527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.19497599601745605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.27200000286102294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.2972800016403198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,power_law_1.2,0.4764671802520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,power_law_1.2,0.49957761764526365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.15815039873123168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.2244352102279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.24622080326080323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.32088959217071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.4250624179840088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.4668543815612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.5111743927001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.5031231880187989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.5152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.5176256179809571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.5431295871734619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.5251327991485596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.5446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.5853312015533447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.602675199508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.5978432178497315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.6213312149047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,power_law_1.01,0.6647295951843262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,power_law_1.01,0.7352128028869629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,power_law_1.01,0.8505663871765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,power_law_1.01,0.8255359649658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,power_law_1.01,0.9231936454772949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,power_law_1.01,1.0912639617919921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,power_law_1.01,1.4162431716918946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,power_law_1.01,1.596332836151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,power_law_1.01,2.1604480743408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,power_law_1.01,2.993152046203613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,power_law_1.01,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,power_law_1.01,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,power_law_1.01,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,power_law_1.01,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,power_law_1.01,0.10434559583663941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,power_law_1.01,0.12311040163040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,power_law_1.01,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,power_law_1.01,0.19048960208892823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,power_law_1.01,0.2664448022842407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,power_law_1.01,0.34117119312286376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.1139456033706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.1523967981338501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.19392640590667726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.24060800075531005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.3072704076766968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.3286400079727173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.34204161167144775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.3429311990737915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.3515199899673462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.36170880794525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.3681983947753906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.37835519313812255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.38670079708099364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.3991616010665894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.42140798568725585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.4664768218994141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5113728046417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6047103881835938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.6699903964996338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.861184024810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.0335424423217774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.395251178741455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,1.7479488372802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,2.46944637298584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,3.184921646118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,power_law_1.2,0.0692736029624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,power_law_1.2,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,power_law_1.2,0.11185920238494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,power_law_1.2,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,power_law_1.2,0.11942399740219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,power_law_1.2,0.12289919853210449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,power_law_1.2,0.12380800247192383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,power_law_1.2,0.12577919960021972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,power_law_1.2,0.12904319763183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,power_law_1.2,0.12739200592041017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,power_law_1.2,0.13400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,power_law_1.2,0.13583359718322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,power_law_1.2,0.1403712034225464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,power_law_1.2,0.15213439464569092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,power_law_1.2,0.17036160230636596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,power_law_1.2,0.1930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,power_law_1.2,0.21870079040527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,power_law_1.2,0.2719615936279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,power_law_1.2,0.31566720008850097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,power_law_1.2,0.43196802139282225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,power_law_1.2,0.5202623844146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,power_law_1.2,0.7140992164611817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,power_law_1.2,0.9232319831848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.09290239810943604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.09579520225524903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.09622399806976319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.09688959717750549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.11434240341186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.12071679830551148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.16016000509262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.17084159851074218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.20753920078277588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,power_law_1.2,1.3148415565490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.26642560958862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.3712064027786255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.4075967788696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.6345791816711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.7846720218658447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.1552767753601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,power_law_1.2,1.6970624923706055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.250387191772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.10268160104751586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.138374400138855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.21722240447998048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.30405120849609374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,power_law_1.2,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,power_law_1.2,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,power_law_1.2,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,power_law_1.2,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,power_law_1.2,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,power_law_1.2,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,power_law_1.2,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,power_law_1.2,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,power_law_1.2,0.12655359506607056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,power_law_1.2,0.1584447979927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,power_law_1.2,0.23979520797729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,power_law_1.2,0.3074048042297363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.037478399276733396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.1144063949584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.14417279958724977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.17600640058517455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.22257280349731445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.318393611907959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.4597760200500488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.3854720115661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.06519039869308471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.11162879467010497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.12435200214385986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.1638208031654358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.20364799499511718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.2535423994064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.3548352003097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.5196479797363281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5497856140136719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,power_law_1.2,0.4290751934051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.6642879962921142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.7152128219604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,power_law_1.2,0.5576511859893799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.13082879781723022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.16895359754562378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.24559359550476073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,power_law_1.2,0.29743359088897703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,power_law_1.2,0.4309120178222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,power_law_1.2,0.6106368064880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,power_law_1.01,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,power_law_1.01,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,power_law_1.01,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,power_law_1.01,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,power_law_1.01,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,power_law_1.01,0.08596479892730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,power_law_1.01,0.08341119885444641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,power_law_1.01,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,power_law_1.01,0.088755202293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,power_law_1.01,0.08790400028228759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,power_law_1.01,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,power_law_1.01,0.10387840270996093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,power_law_1.01,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,power_law_1.01,0.10225919485092164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,power_law_1.01,0.11776000261306763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.01,0.12437759637832642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.01,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.01,0.16040960550308228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.01,0.22012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.01,0.26225919723510743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.01,0.3658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.01,0.5204671859741211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.01,0.6788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.01,0.9233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,balanced,0.05339199801286062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,balanced,0.08768000205357869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,balanced,0.10693333546320598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,balanced,0.10732799768447876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,balanced,0.10605866710344951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,balanced,0.10816533366839091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,balanced,0.10693867007891338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,balanced,0.10641066233317058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,balanced,0.1093386709690094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,balanced,0.10919466614723206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,balanced,0.11036266883214314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,balanced,0.1107413371404012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,balanced,0.11629866560300191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,balanced,0.11552533507347107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,balanced,0.11918399731318156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,balanced,0.1516426702340444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,balanced,0.13708266615867615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,balanced,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,balanced,0.17160000403722128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,balanced,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,balanced,0.21162132422129312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,balanced,0.2673226594924927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,balanced,0.29923733075459796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,balanced,0.408186674118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,balanced,0.4630826711654663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,power_law_1.2,0.025856000185012818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,power_law_1.2,0.02545279860496521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,power_law_1.2,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,power_law_1.2,0.02911359965801239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,power_law_1.2,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,power_law_1.2,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.11946239471435546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.12024320363998413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.18792959451675414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,power_law_1.2,0.2568000078201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,balanced,0.053690666953722634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,balanced,0.055760001142819725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,balanced,0.05633600056171417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,balanced,0.05737066765626272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,balanced,0.06636266907056172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,balanced,0.06622933348019917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,balanced,0.07074133555094402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,balanced,0.09095999598503113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,balanced,0.08738133311271667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,balanced,0.11541866262753804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,balanced,0.12402133146921794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,balanced,0.16247466206550598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,balanced,0.15832533439000449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,balanced,0.20469866196314493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,balanced,0.21595199902852377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,balanced,0.28646934032440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,balanced,0.3354346752166748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,balanced,0.05045866469542185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,balanced,0.07066666583220164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,balanced,0.09569066762924194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,balanced,0.1453386644522349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,balanced,0.24261866013209024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,balanced,0.2420533299446106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,balanced,0.24461867411931357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,balanced,0.24607467651367188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,balanced,0.24702932437260947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,balanced,0.24708267052968344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,balanced,0.2479733427365621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,balanced,0.25093332926432294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,balanced,0.25122666358947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,balanced,0.25693867603937787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,balanced,0.2577226758003235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,balanced,0.26291199525197345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,balanced,0.2723733385403951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,balanced,0.2816213369369507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,balanced,0.29919999837875366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,balanced,0.3203199903170268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,balanced,0.3501013517379761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,balanced,0.37967999776204425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,balanced,0.4693973461786906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,balanced,0.5171999931335449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,balanced,0.7126773198445638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,balanced,0.8046027024586996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,balanced,0.05693866809209188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,balanced,0.059077332417170204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,balanced,0.0605973352988561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,balanced,0.05949866771697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,balanced,0.06311466793219249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,balanced,0.07247466842333476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,balanced,0.07523733377456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,balanced,0.09243733684221904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,balanced,0.1165706713994344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,balanced,0.14146133263905844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,balanced,0.17948265870412192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,balanced,0.19779733816782633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,balanced,0.2762133280436198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,balanced,0.3428106705347697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,balanced,0.4845973253250122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,balanced,0.6157439947128296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,power_law_1.01,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,power_law_1.01,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,power_law_1.01,0.07011200189590454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,power_law_1.01,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,power_law_1.01,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,power_law_1.01,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,power_law_1.01,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,power_law_1.01,0.09380480051040649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,power_law_1.01,0.09873920083045959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,power_law_1.01,0.1080896019935608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,power_law_1.01,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.01,0.13027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.01,0.15430400371551514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.01,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.01,0.2252608060836792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.01,0.27289600372314454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.01,0.3724479913711548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.01,0.47290239334106443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.01,0.7089344024658203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.01,0.8928640365600586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,4,power_law_1.2,0.1126911997795105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,4,power_law_1.2,0.1562432050704956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,4,power_law_1.2,0.17642879486083984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,4,power_law_1.2,0.2137984037399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,4,power_law_1.2,0.25903360843658446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,4,power_law_1.2,0.29261438846588134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,4,power_law_1.2,0.30047359466552737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,4,power_law_1.2,0.3683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,4,power_law_1.2,0.3759488105773926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,4,power_law_1.2,0.3873408079147339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,4,power_law_1.2,0.36002559661865235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,4,power_law_1.2,0.37585279941558836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,4,power_law_1.2,0.36430718898773196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,4,power_law_1.2,0.3684096097946167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,4,power_law_1.2,0.3886591911315918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,4,power_law_1.2,0.40055041313171386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,4,power_law_1.2,0.41591677665710447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,4,power_law_1.2,0.4484543800354004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.2,0.4681663990020752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.2,0.5416128158569335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.2,0.5977024078369141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.2,0.7184127807617188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.2,0.8242943763732911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.2,1.1146368026733398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.2,1.3843775749206544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.2,1.9152128219604492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.2,2.447756767272949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.15112320184707642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.24730238914489747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.43352317810058594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.718611192703247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.0624832153320312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.4724224090576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.2397823333740234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.3264448165893556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.395091247558594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.5076351165771484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.6312255859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.6392831802368164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,power_law_1.2,2.7009727478027346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,power_law_1.2,2.842099189758301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,power_law_1.2,2.9037759780883787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,power_law_1.2,2.9516096115112305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.08287353515625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.313478469848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,power_law_1.2,3.446489715576172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,power_law_1.2,3.866284942626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,power_law_1.2,3.812486267089844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,power_law_1.2,4.296166229248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,power_law_1.2,4.721030426025391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,power_law_1.2,5.7047679901123045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,power_law_1.2,6.210067367553711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,power_law_1.2,8.01580810546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,power_law_1.2,9.309139251708984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,power_law_1.01,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,power_law_1.01,0.11242879629135132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,power_law_1.01,0.14317439794540404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,power_law_1.01,0.1774335980415344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,power_law_1.01,0.23184640407562257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,power_law_1.01,0.24403839111328124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,power_law_1.01,0.2506943941116333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,power_law_1.01,0.25706241130828855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,power_law_1.01,0.26258559226989747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,power_law_1.01,0.27532799243927003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,power_law_1.01,0.278873610496521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,power_law_1.01,0.29142398834228517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,power_law_1.01,0.30410239696502683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,power_law_1.01,0.3071232080459595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,power_law_1.01,0.3209536075592041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,power_law_1.01,0.35359361171722414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,power_law_1.01,0.36896638870239257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,power_law_1.01,0.42753920555114744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,power_law_1.01,0.47219200134277345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,power_law_1.01,0.5814080238342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,power_law_1.01,0.6806015968322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,power_law_1.01,0.8817279815673829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,power_law_1.01,1.0721664428710938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,power_law_1.01,1.4652095794677735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,power_law_1.01,1.8548543930053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.08741120100021363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.1434048056602478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.14163199663162232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.12437119483947753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.12739200592041017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.1285823941230774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.13390079736709595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.13841919898986815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.14942079782485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.14837119579315186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.16017279624938965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.17273600101470948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.17379839420318605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.2014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.2338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.297491192817688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.32622718811035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.4302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.5258944034576416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,0.6977663993835449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,0.9495871543884278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.01,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.01,0.11777280569076538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.01,0.11051520109176635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.01,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.01,0.13776639699935914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.01,0.14506239891052247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.01,0.14629759788513183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.01,0.13891199827194214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.01,0.14458880424499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.01,0.14249600172042848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.01,0.14853119850158691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.01,0.1488703966140747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.01,0.15329279899597167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.01,0.1568511962890625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.01,0.16488959789276122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.01,0.16753920316696166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.01,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.01,0.205401611328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.01,0.21961600780487062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.01,0.26732800006866453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.01,0.29960319995880125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.01,0.36805119514465334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.01,0.4168384075164795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.01,0.5345024108886719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.01,0.6249216079711915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.01,0.9423999786376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.01,1.0637184143066407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.2,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.2,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.2,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.2,0.07812479734420777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.2,0.08852480053901672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.2,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.2,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.2,0.11617920398712159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.2,0.14489599466323852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.2,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.2,0.23237760066986085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.2,0.24746239185333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.2,0.33440001010894777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.2,0.4139711856842041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.2,0.5315711975097657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.2,0.7197184085845947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,power_law_1.01,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,power_law_1.01,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,power_law_1.01,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,power_law_1.01,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,power_law_1.01,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,power_law_1.01,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,power_law_1.01,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,power_law_1.01,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,power_law_1.01,0.06609280109405517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,power_law_1.01,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,power_law_1.01,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,power_law_1.01,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,power_law_1.01,0.10399359464645386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,power_law_1.01,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,power_law_1.01,0.15107200145721436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,power_law_1.01,0.18490240573883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,power_law_1.01,0.2919872045516968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,power_law_1.01,0.357043194770813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,power_law_1.01,0.5972608089447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,power_law_1.01,0.7379072189331055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,power_law_1.01,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,power_law_1.01,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,power_law_1.01,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,power_law_1.01,0.11670399904251098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,power_law_1.01,0.11928319931030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,power_law_1.01,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,power_law_1.01,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,power_law_1.01,0.12986880540847778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,power_law_1.01,0.1322368025779724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,power_law_1.01,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,power_law_1.01,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,power_law_1.01,0.1431872010231018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,power_law_1.01,0.1480512022972107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,power_law_1.01,0.15499520301818848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,power_law_1.01,0.17073919773101806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.01,0.18202879428863525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.01,0.2318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.01,0.27509119510650637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.0931007981300354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.01,0.359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.11442559957504272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.179257595539093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.01,0.4432191848754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.18851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.2010495901107788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.01,0.6104896068572998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.1990847945213318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.20475521087646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.01,0.8539199829101562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.21082239151000975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.21780478954315186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.22264959812164306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.01,1.2073344230651855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.22542080879211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.23551359176635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.24242560863494872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.01,1.6340927124023437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2528127908706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.28199679851531984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.31176319122314455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.3738111972808838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.42378878593444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.5341440200805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.640447998046875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,power_law_1.2,0.849728012084961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.0368639945983886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,power_law_1.2,1.473036766052246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,power_law_1.2,1.8185152053833007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.24540159702301026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.36986238956451417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.5561791896820069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.8254847526550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,power_law_1.2,1.1515839576721192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,power_law_1.2,1.2709055900573731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.7779584884643556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.7030399322509766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,power_law_1.2,1.8089855194091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,power_law_1.2,1.8343040466308593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,power_law_1.2,1.8494527816772461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,power_law_1.2,1.8518335342407226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,power_law_1.2,1.9318016052246094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,power_law_1.2,2.0181312561035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,power_law_1.2,2.0371007919311523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,balanced,0.07965866724650066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,balanced,0.10220266381899516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,balanced,0.14482667048772177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,balanced,0.2291146715482076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,balanced,0.2306399941444397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,balanced,0.23184533913930258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,balanced,0.23262399435043335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,balanced,0.23592533667882284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,balanced,0.23703465859095255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,balanced,0.24470933278401694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,balanced,0.24307199319203696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,balanced,0.24643200635910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,balanced,0.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,balanced,0.25514666239420575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,balanced,0.2626720070838928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,balanced,0.28110400835673016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.1659519195556642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,balanced,0.2966559926668803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,balanced,0.3326080044110616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,balanced,0.36688001950581867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,balanced,0.44255999724070233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,balanced,0.5086880127588908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.283180809020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,balanced,0.6957066853841146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,balanced,0.8161760171254476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,balanced,1.1820106506347656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,power_law_1.2,2.564339256286621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,balanced,1.4795786539713542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,power_law_1.2,2.7218624114990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,power_law_1.2,2.825222396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,power_law_1.2,3.321593475341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,power_law_1.2,3.196339225769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,power_law_1.2,3.6645057678222654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,power_law_1.2,4.483622360229492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,power_law_1.2,5.411718368530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,power_law_1.2,6.3108478546142575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,balanced,0.06564799944559734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,balanced,0.0718506673971812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,balanced,0.07241066793600719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,balanced,0.07501866420110066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,balanced,0.0729013333717982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,balanced,0.08076266447703044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,balanced,0.08080000181992848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,balanced,0.08216000099976857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,balanced,0.07939200103282928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,balanced,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,balanced,0.09064533313115437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,balanced,0.0927946666876475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,balanced,0.09429867068926494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,balanced,0.09924266735712688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,balanced,0.10418132940928142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,balanced,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,balanced,0.12409599622090657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,balanced,0.15647466977437338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,balanced,0.17443199952443442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,balanced,0.23224000136057535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,balanced,0.2772960066795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,balanced,0.3664426803588867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,balanced,0.482474684715271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,balanced,0.05161599814891815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,balanced,0.07502399881680806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,balanced,0.10616000493367513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,balanced,0.11575999855995178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,balanced,0.16383467117945352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,balanced,0.1923253337542216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.07032319903373718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.07971839904785157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.1242624044418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.13556480407714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.16115200519561768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.24113919734954833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.2962752103805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.4282495975494385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.45380477905273436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.7451263904571533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.9933183670043946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,balanced,0.044768000642458596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,balanced,0.07004266480604808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,balanced,0.09796266754468282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,balanced,0.14761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,balanced,0.14989333351453146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,balanced,0.1493119994799296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,balanced,0.15052800377209982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,balanced,0.15030399958292642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,balanced,0.15217600266138712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,balanced,0.15360533197720846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,balanced,0.15250666936238608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,balanced,0.15729066729545593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,balanced,0.16157866517702738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,balanced,0.16475199659665427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,balanced,0.16992000738779703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,balanced,0.17922665675481161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,balanced,0.18758932749430338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,balanced,0.2075200080871582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,balanced,0.2336639960606893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,balanced,0.26710400978724164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,balanced,0.3037866751352946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,balanced,0.39312533537546795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,balanced,0.4673333168029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,balanced,0.6510719855626425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,balanced,0.7822399934132894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,power_law_1.01,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,power_law_1.01,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,power_law_1.01,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,power_law_1.01,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,power_law_1.01,0.1014847993850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,power_law_1.01,0.10124160051345825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,power_law_1.01,0.1018623948097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,power_law_1.01,0.1023360013961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,power_law_1.01,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,power_law_1.01,0.104038405418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,power_law_1.01,0.10826879739761353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,power_law_1.01,0.11615999937057495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,power_law_1.01,0.11747839450836181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,power_law_1.01,0.12992000579833984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,power_law_1.01,0.14561920166015624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.01,0.15815039873123168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.01,0.1838719964027405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.01,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.01,0.2849855899810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.01,0.3456063985824585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.01,0.49728641510009763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.01,0.6638271808624268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.01,1.0544447898864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.01,1.357043170928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.13226879835128785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.15966720581054689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.16699520349502564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.21326079368591308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.2134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.22101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.2527935981750488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.22432639598846435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.24314239025115966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.2479935884475708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.23800959587097167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.2514303922653198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.24288640022277833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.3166271924972534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.37091200351715087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.4092735767364502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.3756799936294556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,power_law_1.01,0.46380162239074707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,power_law_1.01,0.5049215793609619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,power_law_1.01,0.6925119876861572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,power_law_1.01,0.7962175846099854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,power_law_1.01,1.0511360168457031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,power_law_1.01,1.3049856185913087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,power_law_1.01,2.3528959274291994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,power_law_1.01,2.592812728881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,power_law_1.01,4.436959838867187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,power_law_1.01,5.67828483581543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,power_law_1.2,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,power_law_1.2,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,power_law_1.2,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,power_law_1.2,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,power_law_1.2,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,power_law_1.2,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,power_law_1.2,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,power_law_1.2,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,power_law_1.2,0.10098559856414795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,power_law_1.2,0.11779839992523193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,power_law_1.2,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,power_law_1.2,0.19275519847869874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,power_law_1.2,0.24291839599609374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,power_law_1.2,0.28995840549468993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,power_law_1.2,0.42451839447021483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,power_law_1.2,0.5303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,power_law_1.2,0.8145088195800781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,power_law_1.2,0.9623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,balanced,0.07706133524576823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,balanced,0.11003733674685161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,balanced,0.17569067080815634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,balanced,0.2982933322588603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,balanced,0.5502346754074097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,balanced,0.550442655881246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,balanced,0.5526080131530762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,balanced,0.5511786540349325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,balanced,0.5566240151723226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,balanced,0.5568426847457886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,balanced,0.5605119864145914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,balanced,0.5617119868596395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,balanced,0.5676000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,balanced,0.5717600186665853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,balanced,0.5768266518910726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,balanced,0.5857386589050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,balanced,0.59989333152771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,balanced,0.6136746803919474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,balanced,0.6426773468653361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,balanced,0.696885347366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,balanced,0.7730613549550375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,balanced,0.8123733202616373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,balanced,0.9725600083669027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,balanced,1.0307679971059163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,balanced,1.4097760518391926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,balanced,1.591658592224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,power_law_1.01,0.027167999744415285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,power_law_1.01,0.027136000990867614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,power_law_1.01,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,power_law_1.01,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,power_law_1.01,0.03056640028953552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,power_law_1.01,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,power_law_1.01,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,power_law_1.01,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,power_law_1.01,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,power_law_1.01,0.11178879737854004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,power_law_1.01,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,power_law_1.01,0.2001215934753418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,power_law_1.01,0.2763263940811157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,power_law_1.01,0.4232319831848145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,power_law_1.01,0.4932415962219238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.04647040069103241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.14528640508651733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.16195199489593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.22787840366363527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.30672640800476075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.3810111999511719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.01,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.01,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.01,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.01,0.05437440276145935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.01,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.01,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.01,0.08954880237579346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.01,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.01,0.11815040111541748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.01,0.14588799476623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.01,0.18490240573883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.01,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.01,0.31042559146881105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.01,0.37384960651397703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,power_law_1.01,0.4780223846435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.09177600145339966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.11066880226135253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.12848639488220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.19200639724731444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.1978432059288025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.2015552043914795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.2040191888809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.2088383913040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.21742079257965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.22276480197906495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.2310336112976074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.2391808032989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.2509376049041748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.27144320011138917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.3179008007049561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.35809280872344973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.4481088161468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.5177472114562989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.6902783870697021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,0.8476544380187988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,1.1675264358520507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,1.5662976264953614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,2.206188774108887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,2.8076608657836912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.10083839893341065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.12076159715652465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.14680320024490356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.17326079607009887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.21779201030731202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.2275712013244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.23926401138305664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.24138879776000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.24963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.2545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.2582783937454224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.27245440483093264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.28091518878936766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.2950079917907715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.3138432025909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.3684351921081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.40458879470825193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.49710721969604493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.5757567882537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.7499008178710938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,0.9179391860961914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.2590656280517578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,1.6067007064819336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,2.2858367919921876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,3.046240043640137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,power_law_1.2,0.05971199870109558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,power_law_1.2,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,power_law_1.2,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,power_law_1.2,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,power_law_1.2,0.06922240257263183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,power_law_1.2,0.10396800041198731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,power_law_1.2,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,power_law_1.2,0.1566912055015564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,power_law_1.2,0.2069632053375244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,power_law_1.2,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,power_law_1.2,0.32483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,power_law_1.2,0.4194943904876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,power_law_1.2,0.7052608013153077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,power_law_1.2,0.8109312057495117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,power_law_1.2,0.028646400570869444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,power_law_1.2,0.02871040105819702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,power_law_1.2,0.027379199862480164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,power_law_1.2,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,power_law_1.2,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,power_law_1.2,0.04170239865779877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,power_law_1.2,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,power_law_1.2,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.10177919864654542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.16213760375976563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.20599679946899413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.2525631904602051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,power_law_1.2,0.3282111883163452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,power_law_1.01,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,power_law_1.01,0.09057279825210571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,power_law_1.01,0.11597440242767335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,power_law_1.01,0.1397503972053528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,power_law_1.01,0.16751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,power_law_1.01,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,power_law_1.01,0.2068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,power_law_1.01,0.22138240337371826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,power_law_1.01,0.2264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,power_law_1.01,0.23304319381713867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,power_law_1.01,0.23661439418792723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,power_law_1.01,0.24591999053955077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,power_law_1.01,0.254150390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,power_law_1.01,0.265011191368103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,power_law_1.01,0.27930240631103515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,power_law_1.01,0.29310081005096433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,power_law_1.01,0.3259583950042725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,power_law_1.01,0.346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,power_law_1.01,0.41651201248168945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,power_law_1.01,0.46793599128723146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,power_law_1.01,0.6187392234802246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,power_law_1.01,0.7164095878601074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,power_law_1.01,0.9540351867675781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,power_law_1.01,1.168172836303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,power_law_1.01,1.7297023773193358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,power_law_1.01,2.032249641418457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,power_law_1.2,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,power_law_1.2,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,power_law_1.2,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,power_law_1.2,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,power_law_1.2,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,power_law_1.2,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,power_law_1.2,0.14897919893264772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,power_law_1.2,0.1556607961654663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,power_law_1.2,0.16264959573745727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,power_law_1.2,0.1654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,power_law_1.2,0.1792896032333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,power_law_1.2,0.1848896026611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,power_law_1.2,0.18887679576873778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,power_law_1.2,0.20830719470977782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,power_law_1.2,0.213919997215271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,power_law_1.2,0.2178623914718628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,power_law_1.2,0.21965439319610597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,power_law_1.2,0.2513472080230713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,power_law_1.2,0.2730304002761841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,power_law_1.2,0.3261823892593384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,power_law_1.2,0.37587199211120603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,power_law_1.2,0.47543678283691404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,power_law_1.2,0.5535615921020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,power_law_1.2,0.7315648078918457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,power_law_1.2,0.9070464134216308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,power_law_1.2,1.2602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,power_law_1.2,1.6212032318115235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.05514879822731018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.06664320230484008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.07839999794960022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.10054399967193603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.13035520315170288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.13457920551300048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.15690879821777343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.2003391981124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.27565441131591795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.4033535957336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,power_law_1.01,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,power_law_1.01,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,power_law_1.01,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,power_law_1.01,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,power_law_1.01,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,power_law_1.01,0.06510080099105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,power_law_1.01,0.10657919645309448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.5220032215118409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,power_law_1.01,0.13463679552078248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,power_law_1.01,0.17486720085144042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.7036416053771972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,power_law_1.01,0.19478399753570558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,power_law_1.01,0.27757439613342283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,power_law_1.01,0.35771520137786866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,power_law_1.01,0.522764778137207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,power_law_1.01,0.6523712158203125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,balanced,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,balanced,0.06689600149790446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,balanced,0.06614933411280315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,balanced,0.06675200164318085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,balanced,0.0682773341735204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,balanced,0.06881600121657054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,balanced,0.0697866678237915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,balanced,0.07157866656780243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,balanced,0.07207466661930084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,balanced,0.07745600243409474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,balanced,0.07818666597207387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,balanced,0.08228266735871632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,balanced,0.09117866555849712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,balanced,0.10441066821416219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,balanced,0.11348266402880351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,balanced,0.13595733046531677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,balanced,0.15928533673286438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,balanced,0.2038559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,balanced,0.24154667059580484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,balanced,0.34112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,balanced,0.4166026512781779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.07906559705734253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.09251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.11169919967651368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.1438912034034729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.17529599666595458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.2405951976776123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.29071359634399413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.38137600421905515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.48295040130615235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.7085504055023193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.857260799407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,power_law_1.2,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,power_law_1.2,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,power_law_1.2,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.2,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.2,0.10238720178604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.2,0.14068479537963868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.2,0.20129919052124023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.2,0.2425152063369751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.2,0.3676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.2,0.4862271785736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.2,0.8008768081665039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,balanced,0.029343999922275543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,balanced,0.03180266668399175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,balanced,0.03508266558249792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,balanced,0.04083200047413508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,balanced,0.11001599828402202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,balanced,0.1511679987112681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,balanced,0.06503466765085857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,balanced,0.08767466743787129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,balanced,0.10873066385587056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,balanced,0.10912000139554341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,balanced,0.10754666725794475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,balanced,0.10981866717338562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,balanced,0.10997866590817769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,balanced,0.11147200067838033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,balanced,0.11079466342926025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,balanced,0.11370666821797688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,balanced,0.12072533369064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,balanced,0.12167466680208842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,balanced,0.1283573309580485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,balanced,0.13582932949066162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,balanced,0.14808000127474466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,balanced,0.20011732975641885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,balanced,0.19009600083033243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,balanced,0.26070932547251385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,balanced,0.2917226751645406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,balanced,0.4124373197555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,balanced,0.1827253301938375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,balanced,0.2534346580505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,balanced,0.310698668162028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,balanced,0.4814560015996297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,balanced,0.6971893310546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,balanced,0.8605546951293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.2,1.1795392036437988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,power_law_1.2,0.10810879468917847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,power_law_1.2,0.1152575969696045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,power_law_1.2,0.1480831980705261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,power_law_1.2,0.21139199733734132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,power_law_1.2,0.2108288049697876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,power_law_1.2,0.2142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,power_law_1.2,0.22091519832611084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,power_law_1.2,0.2220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,power_law_1.2,0.21709439754486085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,power_law_1.2,0.23285119533538817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,power_law_1.2,0.24019839763641357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,power_law_1.2,0.24650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,power_law_1.2,0.24987521171569824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,power_law_1.2,0.2838848114013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,power_law_1.2,0.31328639984130857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,power_law_1.2,0.3388416051864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,power_law_1.2,0.3075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,power_law_1.2,0.3614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,power_law_1.2,0.45665922164916994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,power_law_1.2,0.5827136039733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,power_law_1.2,0.7491007804870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,power_law_1.2,0.8592896461486816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,power_law_1.2,1.1511296272277831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,power_law_1.2,1.5242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,power_law_1.2,2.4684608459472654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,power_law_1.2,3.498444747924805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,balanced,0.05393599967161814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,balanced,0.07576000193754832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,balanced,0.09298666318257649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,balanced,0.09334400296211243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,balanced,0.0937546690305074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,balanced,0.0974826713403066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,balanced,0.09817066788673401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,balanced,0.09809600313504536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,balanced,0.1016373336315155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,balanced,0.10679466525713603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,balanced,0.11087999741236369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,balanced,0.11773866415023804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,balanced,0.1283146639664968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,balanced,0.13829333583513895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,balanced,0.16085867087046304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,balanced,0.18062933286031088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,balanced,0.2230613430341085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,balanced,0.2666026751200358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,balanced,0.3496853510538737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,balanced,0.4421653350194295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,balanced,0.6160373290379842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,balanced,0.7872052987416586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,power_law_1.2,4.0870014190673825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,power_law_1.2,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,power_law_1.2,0.056652802228927615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,power_law_1.2,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,power_law_1.2,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,power_law_1.2,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,power_law_1.2,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,power_law_1.2,0.07159680128097534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,power_law_1.2,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,power_law_1.2,0.1161728024482727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,power_law_1.2,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,power_law_1.2,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,power_law_1.2,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,power_law_1.2,0.26179840564727785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,power_law_1.2,0.3736511945724487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,power_law_1.2,0.44652161598205564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,balanced,0.052560001611709595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,balanced,0.05548266569773356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,balanced,0.06523199876149495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,balanced,0.07830399771531422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,balanced,0.08190399905045827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,balanced,0.08190399905045827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,balanced,0.08085866769154866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,balanced,0.08458667000134786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,balanced,0.08761599659919739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,balanced,0.08901866277058919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,balanced,0.09225599964459737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,balanced,0.09313600262006123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,balanced,0.0956106682618459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,balanced,0.10365333159764607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,balanced,0.11143466830253601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,balanced,0.12660800417264303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,balanced,0.14029332995414734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,balanced,0.18997333447138467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,balanced,0.22686932484308878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,balanced,0.3296266595522563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,balanced,0.4224426746368408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,balanced,0.6067093213399252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,balanced,0.7961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,power_law_1.01,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,power_law_1.01,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,power_law_1.01,0.062636798620224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,power_law_1.01,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,power_law_1.01,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,power_law_1.01,0.06485120058059693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,power_law_1.01,0.07073919773101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,power_law_1.01,0.07816320061683654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,power_law_1.01,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,power_law_1.01,0.09337599873542786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.01,0.12099200487136841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.01,0.14690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.01,0.15722880363464356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.01,0.1983423948287964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.01,0.2613568067550659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.01,0.3286400079727173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.01,0.4225024223327637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.01,0.6338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.01,0.7679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,balanced,0.040218666195869446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,balanced,0.050794666012128196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.0902933379014333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.10919466614723206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.12744533022244772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.1614346702893575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.19223467508951822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.2642986575762431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.3321546713511149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.05270400047302246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.09147520065307617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.09986559748649597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.12071679830551148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.14587520360946654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.1844928026199341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.20512640476226807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.2807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.3700416088104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.5164415836334229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.7275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,power_law_1.2,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,power_law_1.2,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,power_law_1.2,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,power_law_1.2,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,power_law_1.2,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,power_law_1.2,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,power_law_1.2,0.07678080201148987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.2,0.08863360285758973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.2,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.2,0.1256384015083313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.2,0.16920959949493408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.2,0.20624639987945556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.2,0.2937279939651489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.2,0.36129920482635497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.2,0.6043392181396484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.2,0.7643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,balanced,0.08911466598510742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,balanced,0.08769599596659343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,balanced,0.0925439993540446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,balanced,0.09101333220799764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,balanced,0.0939573347568512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,balanced,0.09277333815892537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,balanced,0.09242133299509685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,balanced,0.0949173370997111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,balanced,0.09383466839790344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,balanced,0.09496000409126282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,balanced,0.09989333152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,balanced,0.10116266210873921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,balanced,0.11729600032170613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,balanced,0.12544533610343933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,balanced,0.10800000031789143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,balanced,0.129013329744339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,balanced,0.14426133036613464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,balanced,0.1521013379096985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,balanced,0.17677332957585654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,balanced,0.20548266172409058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,balanced,0.2980586687723796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,balanced,0.3568906784057617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,balanced,0.5060266653696696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,balanced,0.6085653305053711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,balanced,0.9285333156585693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,balanced,1.1749813556671143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.09972479939460754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.13177599906921386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.18935680389404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.23215360641479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.32193920612335203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.40526719093322755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.4473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.5834432125091553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.5992320060729981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.6350783824920654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,0.6380928039550782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,0.6393792152404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,0.6861440181732178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,0.692467212677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,0.7165887832641602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,0.7070400238037109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,0.7360767841339111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,0.7697279930114747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,0.8065535545349121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,0.8918463706970214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,0.9644479751586914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.171679973602295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,1.2434240341186524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,1.5737664222717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,1.8903615951538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,2.5513919830322265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,3.217452621459961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,balanced,0.03323733309904734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,balanced,0.032629333436489105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,balanced,0.03270933280388514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,balanced,0.04638933142026266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,balanced,0.04619200030962626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,balanced,0.0869760016600291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,balanced,0.10310932993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,balanced,0.13852266470591226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,balanced,0.16914665699005127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,balanced,0.23280533154805502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,balanced,0.2797546585400899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,balanced,0.1370186706384023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,balanced,0.1421173314253489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,balanced,0.14705066879590353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,balanced,0.16545599699020386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,balanced,0.20665599902470908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,balanced,0.27590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,balanced,0.2773066759109497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,balanced,0.2817759911219279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,balanced,0.2826293309529622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,balanced,0.2862880031267802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,balanced,0.2900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,balanced,0.29360000292460126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,balanced,0.2963520089785258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,balanced,0.3039626677831014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,balanced,0.2988106608390808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,balanced,0.3058026631673177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,balanced,0.3162933389345805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,balanced,0.33977067470550537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,balanced,0.36720534165700275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,balanced,0.4275466601053874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,balanced,0.49103466669718426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,1,balanced,0.5754079818725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,1,balanced,0.6775680383046468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,1,balanced,0.9754666487375895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,1,balanced,1.1905919710795085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,1,balanced,1.704426606496175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,1,balanced,2.2292960484822593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.14430079460144044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.144268798828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.14528640508651733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.1491328001022339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.15664639472961425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.1615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.1676095962524414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.17377279996871947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.18320640325546264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.19188480377197265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.22085120677947997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.2437376022338867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.28933761119842527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.3324287891387939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.44046721458435056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.4953343868255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.6910272121429444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.8500288009643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.1962880134582519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,1.4746944427490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,power_law_1.01,0.10332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,power_law_1.01,0.15291520357131957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,power_law_1.01,0.24953598976135255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,power_law_1.01,0.31665279865264895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,power_law_1.01,0.4939839839935303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,power_law_1.01,0.6698624134063721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,power_law_1.01,0.6989952087402344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,power_law_1.01,0.9474687576293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,power_law_1.01,0.9118911743164062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,power_law_1.01,0.8827903747558594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,power_law_1.01,0.9846464157104492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,power_law_1.01,0.9554176330566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,power_law_1.01,1.030790424346924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,power_law_1.01,1.0026495933532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,power_law_1.01,1.0545856475830078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,power_law_1.01,1.0734911918640138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,power_law_1.01,1.133420753479004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,power_law_1.01,1.1409215927124023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,power_law_1.01,1.253990364074707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,power_law_1.01,1.364243221282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,power_law_1.01,1.4645183563232422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,power_law_1.01,1.6540864944458007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,power_law_1.01,1.7448896408081054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,power_law_1.01,2.0918399810791017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,power_law_1.01,2.4647424697875975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,power_law_1.01,3.1018367767333985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,power_law_1.01,3.8046398162841797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,balanced,0.05235733091831207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,balanced,0.055567999680837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,balanced,0.06889600058396657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,balanced,0.06784533460934956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,balanced,0.07080533107121785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,balanced,0.0730560024579366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,balanced,0.07287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,balanced,0.08380267024040222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,balanced,0.0869813362757365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,balanced,0.09623466928799947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.10214933753013611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.1151093343893687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.136762668689092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.15851199626922607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.18338133891423544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.24073066314061484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.2911786635716756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.4039306640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.5125600099563599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,power_law_1.2,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,power_law_1.2,0.10496640205383301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,power_law_1.2,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,power_law_1.2,0.12483199834823608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,power_law_1.2,0.12070399522781372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,power_law_1.2,0.12212480306625366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,power_law_1.2,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,power_law_1.2,0.12874879837036132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,power_law_1.2,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,power_law_1.2,0.12395520210266113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,power_law_1.2,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,power_law_1.2,0.12814719676971437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,power_law_1.2,0.1443392038345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,power_law_1.2,0.1428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,power_law_1.2,0.14842239618301392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,power_law_1.2,0.16405119895935058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.2,0.1706112027168274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.2,0.2178112030029297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.2,0.25809919834136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.2,0.33245439529418946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.2,0.3981184005737305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.2,0.5476607799530029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.2,0.7464960098266602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.2,1.051027202606201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.2,1.446617603302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,power_law_1.01,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,power_law_1.01,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,power_law_1.01,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,power_law_1.01,0.10383360385894776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,power_law_1.01,0.10892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,power_law_1.01,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,power_law_1.01,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,power_law_1.01,0.11482239961624145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,power_law_1.01,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,power_law_1.01,0.1184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,power_law_1.01,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,power_law_1.01,0.12438399791717529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,power_law_1.01,0.1303104043006897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,power_law_1.01,0.13054720163345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,power_law_1.01,0.13305599689483644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,power_law_1.01,0.14776320457458497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.1911936044692993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.24988160133361817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.3171328067779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.39377920627593993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,power_law_1.01,0.4688000202178955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,power_law_1.01,0.6546048164367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,power_law_1.01,0.8055808067321777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.12817920446395875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.17637759447097778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.18883199691772462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.18954240083694457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.19246079921722412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.21329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.21176960468292236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.21710081100463868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.23588480949401855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.24631040096282958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.26023681163787843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.295142388343811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.3050559997558594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.3614464044570923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.426636791229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.5564608097076416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.6289984226226807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,power_law_1.2,0.821555233001709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.086911964416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,power_law_1.2,1.461689567565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,power_law_1.2,1.819968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,power_law_1.2,0.05575039982795715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,power_law_1.2,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,power_law_1.2,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,power_law_1.2,0.1253056049346924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,power_law_1.2,0.18343679904937743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,power_law_1.2,0.1876863956451416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,power_law_1.2,0.19082239866256714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,power_law_1.2,0.19555840492248536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,power_law_1.2,0.191103994846344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,power_law_1.2,0.2098560094833374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,power_law_1.2,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,power_law_1.2,0.21802880764007568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,power_law_1.2,0.22344319820404052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,power_law_1.2,0.23447039127349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,power_law_1.2,0.24247679710388184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,power_law_1.2,0.24915199279785155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,power_law_1.2,0.2844288110733032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,power_law_1.2,0.3023808002471924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,power_law_1.2,0.36138238906860354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,power_law_1.2,0.4035520076751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,power_law_1.2,0.5269824028015136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,power_law_1.2,0.5985856056213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,power_law_1.2,0.8072447776794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,power_law_1.2,1.041055965423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,power_law_1.2,1.335244846343994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,power_law_1.2,1.881158447265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,balanced,0.038831998904546104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,balanced,0.03262399882078171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,balanced,0.036373332142829895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,balanced,0.06991999844710033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,balanced,0.10428800185521443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,power_law_1.2,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,power_law_1.2,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,power_law_1.2,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,power_law_1.2,0.16974079608917236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,power_law_1.2,0.17968640327453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,power_law_1.2,0.1848512053489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,power_law_1.2,0.1889024019241333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,power_law_1.2,0.201964807510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,power_law_1.2,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,power_law_1.2,0.22449920177459717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,power_law_1.2,0.24045441150665284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,power_law_1.2,0.26098558902740476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,power_law_1.2,0.2730623960494995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,power_law_1.2,0.26472320556640627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,power_law_1.2,0.3244415998458862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,power_law_1.2,0.3639872074127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,power_law_1.2,0.4558080196380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,power_law_1.2,0.5326144218444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,power_law_1.2,0.6994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,power_law_1.2,0.8684288024902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,power_law_1.2,1.180742359161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,power_law_1.2,1.5030271530151367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,power_law_1.2,2.122777557373047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,power_law_1.2,2.787571144104004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,balanced,0.08214400211970012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,balanced,0.09514666597048442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,balanced,0.11220266421635945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,balanced,0.11416000127792358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,balanced,0.1111306647459666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,balanced,0.11381333072980244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,balanced,0.11427733302116394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,balanced,0.11496000488599141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,balanced,0.11712533235549927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,balanced,0.11708799997965495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,balanced,0.12614400188128153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,balanced,0.12811733285586038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,balanced,0.13450666268666586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,balanced,0.14075733224550882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,balanced,0.15003732840220133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,balanced,0.16660267114639282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,balanced,0.18015466133753458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,balanced,0.2205280065536499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,balanced,0.2523039976755778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,balanced,0.32547734181086224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,balanced,0.42634133497873944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,balanced,0.5843679904937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,balanced,0.778602679570516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,balanced,0.037962667644023895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,balanced,0.06788266698519389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,balanced,0.06756266454855601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,balanced,0.06714666883150737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,balanced,0.06955199937025706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,balanced,0.06886399785677592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,balanced,0.08170666793982188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,balanced,0.08447999755541484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.10066133737564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.10757866501808167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.11364799737930298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.13770133256912231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.16359466314315796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.21081066131591797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.2430986762046814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.33399466673533124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.42342400550842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,power_law_1.01,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,power_law_1.01,0.08596479892730713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,power_law_1.01,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,power_law_1.01,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,power_law_1.01,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,power_law_1.01,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,power_law_1.01,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,power_law_1.01,0.10289280414581299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,power_law_1.01,0.11415040493011475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,power_law_1.01,0.11706880331039429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,power_law_1.01,0.12922240495681764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,power_law_1.01,0.136518394947052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,power_law_1.01,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.01,0.25771520137786863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.01,0.3181567907333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.01,0.3878655910491943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.01,0.4936831951141357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.01,0.9155648231506348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.01,1.1195136070251466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.01,1.5283519744873046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.01,2.1770496368408203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.01,2.7486143112182617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,balanced,0.07566933333873749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,balanced,0.07464533547560374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,balanced,0.07729599873224895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,balanced,0.07631466786066692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,balanced,0.07709866762161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,balanced,0.07763200004895528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,balanced,0.0788213312625885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,balanced,0.0804746647675832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,balanced,0.08092266817887624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,balanced,0.08213866750399272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,balanced,0.09031466643015544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,balanced,0.0962506632010142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,balanced,0.09970133503278096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,balanced,0.10074133674303691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,balanced,0.11257066329320271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,balanced,0.12744533022244772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,balanced,0.15495466192563376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,balanced,0.18913600842158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,balanced,0.24572267134984335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,balanced,0.2890080014864604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,balanced,0.3978186845779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,balanced,0.4851413170496623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,balanced,0.7059733072916666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,balanced,0.9052159786224365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,balanced,0.10718933741251628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,balanced,0.1732906699180603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,balanced,0.2601439952850342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,balanced,0.46960000197092694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,balanced,0.8787679672241211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,balanced,1.093664010365804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,balanced,1.0966026782989502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,balanced,1.1011573473612468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,balanced,1.1074613730112712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,balanced,1.1034399668375652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,balanced,1.104149341583252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,balanced,1.1125600337982178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,balanced,1.1162880261739094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,balanced,1.1241546471913655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,balanced,1.1399359703063965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,balanced,1.1484959920247395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,balanced,1.1571199893951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,balanced,1.239194631576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,balanced,1.203978697458903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,balanced,1.5358667373657227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,balanced,1.3182400067647297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,balanced,2.0135092735290527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,balanced,1.562394618988037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,balanced,2.4725119272867837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,balanced,2.327669302622477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,balanced,3.9120639165242515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,balanced,4.143589337666829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,power_law_1.01,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,power_law_1.01,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,power_law_1.01,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,power_law_1.01,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,power_law_1.01,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,power_law_1.01,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,power_law_1.01,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.01,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.01,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.01,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.01,0.12709120512008668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.01,0.17734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.01,0.22501120567321778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.01,0.299564790725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.01,0.39130239486694335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,power_law_1.01,0.06529920101165772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,power_law_1.01,0.0625216007232666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,power_law_1.01,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,power_law_1.01,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,power_law_1.01,0.07443199753761291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,power_law_1.01,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,power_law_1.01,0.07525759935379028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,power_law_1.01,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,power_law_1.01,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,power_law_1.01,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,power_law_1.01,0.10391680002212525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,power_law_1.01,0.10734080076217652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,power_law_1.01,0.1328511953353882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,power_law_1.01,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,power_law_1.01,0.1935807943344116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,power_law_1.01,0.2488703966140747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,power_law_1.01,0.3137343883514404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,power_law_1.01,0.43187198638916013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,power_law_1.01,0.5563136100769043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,balanced,0.03418133407831192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,balanced,0.03818666686614355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,balanced,0.04080000023047129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,balanced,0.043477331598599754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,balanced,0.05861333509286245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,power_law_1.01,0.08012800216674805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,balanced,0.09598400195439656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,balanced,0.11045333743095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,balanced,0.1421013375123342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,power_law_1.01,0.08732799887657165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,balanced,0.17097600301106772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,balanced,0.24838932355244955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,power_law_1.01,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,balanced,0.32014399766921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,power_law_1.01,0.15822720527648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,power_law_1.01,0.18275840282440187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,power_law_1.01,0.23724160194396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,power_law_1.01,0.23372159004211426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,power_law_1.01,0.24289920330047607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,power_law_1.01,0.2551232099533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,power_law_1.01,0.2638144016265869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,power_law_1.01,0.26375679969787597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,power_law_1.01,0.2702080011367798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,power_law_1.01,0.2762239933013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,power_law_1.01,0.2860543966293335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,power_law_1.01,0.29190399646759035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,power_law_1.01,0.3030848026275635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,power_law_1.01,0.3416064023971558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,power_law_1.01,0.34235520362854005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,power_law_1.01,0.3882431983947754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,power_law_1.01,0.40967679023742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,power_law_1.01,0.4983424186706543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,power_law_1.01,0.5320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,power_law_1.01,0.7045631885528565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,power_law_1.01,0.8057408332824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,power_law_1.01,1.1146240234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,power_law_1.01,1.4134528160095214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,power_law_1.2,0.09156479835510253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,power_law_1.2,0.09091200232505799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,power_law_1.2,0.10678399801254272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,power_law_1.2,0.12116479873657227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,power_law_1.2,0.12128000259399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,power_law_1.2,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,power_law_1.2,0.1499071955680847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,power_law_1.2,0.14128639698028564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,power_law_1.2,0.1443519949913025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,power_law_1.2,0.14632960557937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,power_law_1.2,0.15232640504837036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,power_law_1.2,0.15310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,power_law_1.2,0.16128000020980834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,power_law_1.2,0.16589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,power_law_1.2,0.166975998878479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,power_law_1.2,0.17402880191802977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.2,0.21084799766540527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.2,0.2748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.2,0.3150464057922363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.2,0.4182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.2,0.5369664192199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.2,0.7284095764160157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.2,0.9548864364624023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.2,1.3536255836486817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.2,1.9411008834838868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,power_law_1.01,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,power_law_1.01,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,power_law_1.01,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,power_law_1.01,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,power_law_1.01,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,power_law_1.01,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.01,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.01,0.09504640102386475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.01,0.10680320262908935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.01,0.1371072053909302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.01,0.16436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.01,0.2267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.01,0.2818304061889648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.01,0.39403519630432127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.01,0.5326144218444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,power_law_1.2,0.077811199426651
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,power_law_1.2,0.07810559868812561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,power_law_1.2,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,power_law_1.2,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,power_law_1.2,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,power_law_1.2,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,power_law_1.2,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,power_law_1.2,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,power_law_1.2,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,power_law_1.2,0.09292799830436707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,power_law_1.2,0.09934719800949096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,power_law_1.2,0.10156160593032837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,power_law_1.2,0.11034239530563354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,power_law_1.2,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,power_law_1.2,0.1375872015953064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,power_law_1.2,0.17242239713668822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,power_law_1.2,0.20874240398406982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,power_law_1.2,0.2592895984649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,power_law_1.2,0.31332480907440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,power_law_1.2,0.4093952178955078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,balanced,0.037231999138991036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,balanced,0.04008533308903376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,balanced,0.043477331598599754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,balanced,0.0890773336092631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,balanced,0.09020266930262248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,balanced,0.11802132924397786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,balanced,0.12989333271980286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,balanced,0.17160000403722128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,balanced,0.20416533946990967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,balanced,0.29595200220743817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,balanced,0.3737866481145223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,balanced,0.5417493184407552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,balanced,0.6977333227793375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,power_law_1.2,0.6047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,power_law_1.2,0.7885759830474853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,power_law_1.2,1.0125184059143066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.01,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.01,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.01,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.01,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.01,0.10744320154190064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.01,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.01,0.1386944055557251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.01,0.12508800029754638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.01,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.01,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.01,0.20953600406646727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.01,0.2586496114730835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.01,0.37317121028900146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.01,0.3711103916168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,power_law_1.01,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,power_law_1.01,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,power_law_1.01,0.043263998627662656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,power_law_1.01,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,power_law_1.01,0.05681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,power_law_1.01,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.01,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.01,0.15649280548095704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.01,0.17352319955825807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.01,0.26673920154571534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.01,0.35167360305786133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.01,0.5520768165588379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.01,0.7126272201538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.01,1.0773759841918946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.01,1.4904383659362792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.2,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.2,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.2,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.2,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.2,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.2,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.2,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.2,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.2,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.2,0.12474240064620971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.2,0.16133760213851928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.2,0.18573440313339235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.2,0.2744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.2,0.34980480670928954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.2,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.2,0.06347519755363465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.2,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.2,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.2,0.0695743978023529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.2,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.2,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.2,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.2,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.2,0.08758400082588196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.2,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.2,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.2,0.13025920391082763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.2,0.143449604511261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.2,0.1962175965309143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.2,0.2163072109222412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.2,0.3230079889297485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.2,0.39549438953399657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.2,0.5151167869567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.2,0.6695551872253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,balanced,0.05625066657861074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,balanced,0.11130666732788086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,balanced,0.17325866222381592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,balanced,0.17689067125320435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,balanced,0.18330132961273193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,balanced,0.18077866236368814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,balanced,0.18422933419545492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,balanced,0.18996800978978476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,balanced,0.1846346656481425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,balanced,0.19322667519251505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,balanced,0.19748800992965698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,balanced,0.20302400986353555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,balanced,0.20721065998077393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,balanced,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,balanced,0.22393600145975748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,balanced,0.2539520064989726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.2804746627807617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.33532265822092694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.38499732812245685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.5344213247299194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.6269813378651937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.9013120333353678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,balanced,1.1332746346791585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,balanced,1.6574400266011555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,balanced,2.177610715230306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,power_law_1.2,0.026502400636672974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,power_law_1.2,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,power_law_1.2,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,power_law_1.2,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,power_law_1.2,0.059935998916625974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,power_law_1.2,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,power_law_1.2,0.06959999799728393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,power_law_1.2,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.1387712001800537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.16102399826049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.20331521034240724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.05372160077095032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.06128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.08352640271186829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.12343679666519165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.13232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.154476797580719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.2071552038192749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.25188479423522947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,power_law_1.01,0.31916160583496095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,power_law_1.01,0.5832575798034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,power_law_1.01,0.7251008033752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.29460480213165285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.3733695983886719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,power_law_1.2,0.4619328022003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,power_law_1.01,1.1179583549499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,power_law_1.01,1.2363967895507812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,power_law_1.01,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,power_law_1.01,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,power_law_1.01,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,power_law_1.01,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,power_law_1.01,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,power_law_1.01,0.07711359858512878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,power_law_1.01,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,power_law_1.01,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.01,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.01,0.11606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.01,0.13320319652557372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.01,0.16035200357437135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.01,0.19447040557861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.01,0.273471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.01,0.3450623989105225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.01,0.5173183917999268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.01,0.6706880092620849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,balanced,0.059861332178115845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,balanced,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,balanced,0.08225066463152568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,balanced,0.07849066456158955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,balanced,0.07986133297284444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,balanced,0.07843199868996938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,balanced,0.07863466441631317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,balanced,0.08125866452852885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,balanced,0.0806933343410492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,balanced,0.08166933556397755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,balanced,0.08877866466840108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,balanced,0.1046346624692281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,balanced,0.11498133341471355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,balanced,0.12635733683904013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,balanced,0.15004266301790872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,balanced,0.17765865723292032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,balanced,0.22265066703160605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,balanced,0.2662026683489482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,balanced,0.3617173433303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,balanced,0.4505759874979655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,balanced,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,balanced,0.04109866668780645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,balanced,0.04417600234349569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,balanced,0.0452106644709905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,balanced,0.04820266862710317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,balanced,0.05972800155480703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,balanced,0.08050133287906647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,balanced,0.08147733410199483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,balanced,0.10070400436719258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,balanced,0.1220906674861908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,balanced,0.1581653356552124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,balanced,0.20772266387939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,balanced,0.3007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,balanced,0.3859573205312093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,balanced,0.5555306673049927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,balanced,0.7184533278147379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.13557759523391724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.20202879905700682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.22481920719146728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.4145664215087891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.43082242012023925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.44356479644775393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.45742077827453614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.4488639831542969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.457747220993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.46599678993225097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.4723264217376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.5019904136657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.513260793685913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.5313151836395263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.6040832042694092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.6108992099761963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,power_law_1.2,0.5890111923217773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,power_law_1.2,0.6574592113494873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,power_law_1.2,0.7819583892822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,power_law_1.2,0.9089728355407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,power_law_1.2,1.0832832336425782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,power_law_1.2,1.3836223602294921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,power_law_1.2,1.6876928329467773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,power_law_1.2,2.37142391204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,power_law_1.2,3.133600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,power_law_1.2,4.566284942626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,power_law_1.2,6.0027519226074215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.10236799716949463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.1343808054924011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.14279040098190307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.1482432007789612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.1579200029373169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.15852160453796388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.16547839641571044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.17299840450286866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.17674880027770995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.19139839410781861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.19761919975280762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.19759360551834107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.2297600030899048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2577471971511841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.30084478855133057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.3691711902618408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.4818111896514893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.5492800235748291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.7710464000701904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.0811519622802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,1.403551959991455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,1.8034944534301758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,balanced,0.03516799956560135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,balanced,0.04582933088143667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,balanced,0.07156266768773396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,balanced,0.1313759982585907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,balanced,0.13127999504407248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,balanced,0.12757866581281027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,balanced,0.12965333461761475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,balanced,0.1299679974714915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,balanced,0.1316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,balanced,0.1352000037829081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,balanced,0.13613333304723105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,balanced,0.13592533270517984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,balanced,0.14074666301409403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,balanced,0.14153066277503967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,balanced,0.1448746621608734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,balanced,0.1729173262914022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,balanced,0.16531200210253397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,balanced,0.20989332596460977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,balanced,0.21527999639511108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,balanced,0.30244266986846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,balanced,0.32154132922490436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,balanced,0.43944533665974933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,balanced,0.5061173439025879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,balanced,0.748581329981486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,balanced,0.9186933040618896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,balanced,0.0401653324564298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,balanced,0.06154666841030121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,balanced,0.0867733359336853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,balanced,0.08673600355784099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,balanced,0.08649599552154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,balanced,0.08730666836102803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,balanced,0.08665600419044495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,balanced,0.08839999636014302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,balanced,0.08796800176302592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,balanced,0.09233599901199341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,balanced,0.09102933605511983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,balanced,0.09326933821042378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,balanced,0.09340266386667888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,balanced,0.10876267155011494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,balanced,0.10973866780598958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,balanced,0.10946133732795715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,balanced,0.13109866778055826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,balanced,0.13985600074132284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,balanced,0.1715786655743917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,balanced,0.19300266106923422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,balanced,0.255295991897583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,balanced,0.31862932443618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,balanced,0.44868266582489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,balanced,0.56495467821757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,balanced,0.8272319634755453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,balanced,1.0748480161031086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,power_law_1.01,0.050291198492050174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,power_law_1.01,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,power_law_1.01,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,power_law_1.01,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,power_law_1.01,0.07937920093536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,power_law_1.01,0.09751039743423462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,power_law_1.01,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,power_law_1.01,0.10034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,power_law_1.01,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,power_law_1.01,0.10304640531539917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,power_law_1.01,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,power_law_1.01,0.11235200166702271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,power_law_1.01,0.11541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,power_law_1.01,0.11565439701080323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,power_law_1.01,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,power_law_1.01,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,power_law_1.01,0.14870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,power_law_1.01,0.17423360347747802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,power_law_1.01,0.20373759269714356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,power_law_1.01,0.26343679428100586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,power_law_1.01,0.3384255886077881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,power_law_1.01,0.47660160064697266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,power_law_1.01,0.6379456043243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,power_law_1.01,0.9415167808532715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,power_law_1.01,1.1998592376708985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,balanced,0.031194667021433514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,balanced,0.034287999073664345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,balanced,0.06453333298365276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,balanced,0.08604799707730611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,balanced,0.11379733681678772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,balanced,0.13436266779899597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,balanced,0.17884800831476846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,balanced,0.22390933831532797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.11744639873504639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.12055679559707641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.12398719787597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.1276927947998047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.1294144034385681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.13479039669036866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.13563519716262817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.1482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.15089919567108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.15954560041427612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.19148800373077393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.22186241149902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.2632960081100464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.31967360973358155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.3999295949935913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.4911808013916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.6788928031921386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,power_law_1.2,0.86561279296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.237446403503418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,power_law_1.2,1.52542724609375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.13215359449386596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.17752959728240966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.19752960205078124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,power_law_1.01,0.28054399490356446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,power_law_1.01,0.3711999893188477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,balanced,0.0458186666170756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,balanced,0.04619200030962626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,balanced,0.05778133372465769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.0664213349421819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.07656533519426982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.08273066580295563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11424000064531963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.13519466916720072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.18144534031550089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.20764267444610596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.28786667188008624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.3656959931055705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,power_law_1.2,0.0608959972858429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,power_law_1.2,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,power_law_1.2,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,power_law_1.2,0.1464959979057312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,power_law_1.2,0.18348159790039062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,power_law_1.2,0.19146239757537842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,power_law_1.2,0.19438719749450684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,power_law_1.2,0.19353599548339845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,power_law_1.2,0.2088383913040161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,power_law_1.2,0.2100287914276123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,power_law_1.2,0.21643519401550293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,power_law_1.2,0.22220799922943116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,power_law_1.2,0.23509759902954103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,power_law_1.2,0.24673919677734374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,power_law_1.2,0.2542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,power_law_1.2,0.2778752088546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.2,0.2988352060317993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.2,0.35617918968200685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.2,0.4086463928222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.2,0.4986112117767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.2,0.5941504001617431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.2,0.7838335990905761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.2,0.9735679626464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.2,1.3316672325134278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.2,1.788595199584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.13192960023880004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.13559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.14011520147323608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.1384384036064148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.14277759790420533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.14453760385513306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.1506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.16721919775009156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.16922880411148072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.18698879480361938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.21178879737854003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.2253567934036255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.2767999887466431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.3210047960281372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,power_law_1.2,0.4129983901977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,power_law_1.2,0.501196813583374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,power_law_1.2,0.6907455921173096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,power_law_1.2,0.880742359161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,power_law_1.2,1.2559167861938476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,balanced,0.15693333745002747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,balanced,0.26065067450205487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,balanced,0.4644639889399211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,balanced,0.8776106834411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,balanced,1.7102452913920085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,balanced,2.1300586064656577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,balanced,2.1319573720296225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,balanced,2.1471734046936035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,balanced,2.154714743296305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,balanced,2.1569066047668457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,balanced,2.1644159952799478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,balanced,2.1778027216593423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,balanced,2.183274745941162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,balanced,2.199717362721761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,balanced,2.223562717437744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,balanced,2.243562698364258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,power_law_1.2,1.62410888671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,balanced,2.271519978841146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,balanced,2.2988692919413247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,balanced,2.3479199409484863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,balanced,2.9129387537638345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,balanced,2.5617012977600098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,balanced,3.8932905197143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,balanced,2.980064074198405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,balanced,5.0975039800008135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,balanced,4.4952694574991865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,balanced,7.813178380330403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,balanced,8.061936060587565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,balanced,0.044362664222717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,balanced,0.0558240016301473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,balanced,0.06260266900062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,balanced,0.073594664533933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,balanced,0.08121066788832347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,balanced,0.09299199779828389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,balanced,0.09516800443331401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,balanced,0.11534399787584941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,balanced,0.12760532895723978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,balanced,0.15850133697191873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,balanced,0.18102933963139853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,balanced,0.233952005704244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,balanced,0.2836959958076477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,balanced,0.39293865362803143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,balanced,0.5023200114568075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.13559039831161498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14022400379180908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14481279850006104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.14582400321960448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.15416959524154664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.16329599618911744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.16836479902267457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.17857919931411742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.19040640592575073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.20251519680023194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.24627199172973632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.282041597366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.36229119300842283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.43754239082336427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.5800384044647217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7243135929107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,power_law_1.01,0.9981375694274902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.2697152137756347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.7792959213256836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.2702848434448244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,power_law_1.2,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,power_law_1.2,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,power_law_1.2,0.08550400137901307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,power_law_1.2,0.11308799982070923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,power_law_1.2,0.15062400102615356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,power_law_1.2,0.19456640481948853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,power_law_1.2,0.22376959323883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,power_law_1.2,0.2767168045043945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,power_law_1.2,0.29263360500335694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,power_law_1.2,0.3019968032836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,power_law_1.2,0.31198720932006835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,power_law_1.2,0.33376638889312743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,power_law_1.2,0.34112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,power_law_1.2,0.36550400257110593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,power_law_1.2,0.38472321033477785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,power_law_1.2,0.4053631782531738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,power_law_1.2,0.437011194229126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,power_law_1.2,0.5226431846618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,power_law_1.2,0.5410367965698242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,power_law_1.2,0.6573760032653808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,power_law_1.2,0.729695987701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,power_law_1.2,0.9255871772766113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,power_law_1.2,1.134995174407959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,power_law_1.2,1.5068672180175782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,power_law_1.2,1.9312320709228517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,power_law_1.2,2.625593566894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,power_law_1.2,3.215302276611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.08917120099067688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.09486079812049866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.10917119979858399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.1293503999710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.13692159652709962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.16350719928741456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.18480639457702636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.24479360580444337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.28897919654846194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,power_law_1.01,0.39685759544372556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,power_law_1.01,0.49886717796325686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,balanced,0.0558186670144399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,balanced,0.05656533439954122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,balanced,0.056330665946006775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,balanced,0.065461332599322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,balanced,0.06790400048096974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,balanced,0.07875733574231465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,balanced,0.10294933120409648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,balanced,0.11540266871452332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,balanced,0.15452800194422403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,balanced,0.1580959955851237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,balanced,0.19639466206232706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,balanced,0.22188800573349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,balanced,0.2945920030275981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,balanced,0.353061318397522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,power_law_1.01,0.7063295841217041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,power_law_1.01,0.8936639785766601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,power_law_1.2,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,power_law_1.2,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,power_law_1.2,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,power_law_1.2,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,power_law_1.2,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,power_law_1.2,0.05604479908943176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,power_law_1.2,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.2,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.2,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.2,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.2,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.2,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.2,0.1832576036453247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.2,0.24865920543670655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.2,0.35406079292297366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.2,0.4399360179901123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.06398079991340637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.09416959881782531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.09120640158653259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.10350719690322877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.14602880477905272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.18536959886550902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.24406399726867675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.2989567995071411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.4397568225860596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.5201407909393311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,power_law_1.2,0.75382399559021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,power_law_1.2,0.9081024169921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,balanced,0.06451199948787689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,balanced,0.046351999044418335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,balanced,0.047498668233553566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,balanced,0.05306666592756907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,balanced,0.05834666887919108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,balanced,0.07014933228492737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,balanced,0.09678933024406433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,balanced,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,balanced,0.1495039959748586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,balanced,0.1787733236948649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,balanced,0.24025599161783853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,balanced,0.29739199082056683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,balanced,0.4173653523127238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,balanced,0.5337333281834921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1524608016014099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.23224959373474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.3083584070205688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.35129599571228026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.4636096000671387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.47562241554260254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.49982080459594724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5093632221221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5396416187286377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.5418560028076171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.5636032104492188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.5874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.6164671897888183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.6224512100219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.6951615810394287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.749836778640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.8715840339660644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,power_law_1.01,0.9377471923828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.1356096267700195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.3990655899047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,power_law_1.01,1.7736576080322266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.0550336837768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,power_law_1.01,2.7309759140014647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,power_law_1.01,3.4774143218994142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10196479558944702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.15356800556182862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.21763839721679687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.29589118957519533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.45856637954711915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.4753983974456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5010176181793213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.5106560230255127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5033215999603271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.5268032073974609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.5521728038787842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.5610239982604981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.578937578201294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.596281623840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.6016064167022706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.6219007968902588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.6748479843139649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.7269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,power_law_1.01,0.8356479644775391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,power_law_1.01,0.9367103576660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.1547391891479493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.2446208000183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,power_law_1.01,1.5687487602233887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,power_law_1.01,2.0201791763305663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,power_law_1.01,2.698624038696289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,power_law_1.01,3.5392127990722657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,balanced,0.07732800145943959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,balanced,0.07735999921957652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,balanced,0.07766399780909221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,balanced,0.07971733311812083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,balanced,0.07886399825414021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,balanced,0.07931733131408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,balanced,0.08106133341789246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,balanced,0.08389866352081299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,balanced,0.08541333675384521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,balanced,0.09067199627558391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,balanced,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,balanced,0.10017599662144978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,balanced,0.10379200180371602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,balanced,0.1218986709912618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,balanced,0.13133866588274637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,balanced,0.15692266821861267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,balanced,0.18644267320632935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,balanced,0.24366400639216104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,balanced,0.29757867256800336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,balanced,0.41150931517283124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,balanced,0.526528000831604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,balanced,0.7546986738840739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,balanced,0.9765600363413492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.09086080193519593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.10906239748001098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.1547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.16829440593719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.17436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.1742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.1805888056755066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.18572800159454345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.19326080083847047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.1965440034866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.2068864107131958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.21210238933563233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.22383360862731932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.25982720851898194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.2920383930206299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.3589247941970825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.4162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.5511807918548584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.6766975879669189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,0.9314111709594727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.17706241607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,1.6793920516967773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.09215360283851623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.12611839771270753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,2.1819711685180665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.1607807993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.1996608018875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.2562688112258911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.32763519287109377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.0960319995880127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.09667840003967285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.09728000164031983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.09731839895248413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,power_law_1.2,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,power_law_1.2,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,power_law_1.2,0.03310079872608185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.11761280298233032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11925760507583619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12657920122146607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,power_law_1.2,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,power_law_1.2,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,power_law_1.2,0.053600001335144046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.13842560052871705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.19283839464187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.24764161109924315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.4640768051147461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,power_law_1.2,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,power_law_1.2,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.28958079814910886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.39763200283050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,power_law_1.2,0.17455999851226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.5151040077209472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,power_law_1.2,0.24223361015319825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.7345151901245117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.5985343933105469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.9327743530273438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,power_law_1.2,0.3240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.381382369995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.8138431549072265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.10362880229949951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.10062079429626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.11284480094909669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.10657919645309448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.11260160207748413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.11673599481582642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.11159039735794067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.1283776044845581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.14580479860305787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.1662976026535034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.19678720235824584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.23226239681243896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.28592638969421386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.36357119083404543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.45948162078857424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,power_law_1.01,0.6662784099578858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,power_law_1.01,0.8175488471984863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,power_law_1.01,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,power_law_1.01,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,power_law_1.01,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,power_law_1.01,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,power_law_1.01,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,power_law_1.01,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,power_law_1.01,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,power_law_1.01,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,power_law_1.01,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.01,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.01,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.01,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.01,0.1861248016357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.01,0.2870016098022461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.01,0.3712768077850342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.01,0.6047679901123046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.01,0.9537983894348144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,power_law_1.2,0.1089087963104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,power_law_1.2,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,power_law_1.2,0.1167296051979065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,power_law_1.2,0.11845760345458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,power_law_1.2,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,power_law_1.2,0.12577279806137084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,power_law_1.2,0.1281216025352478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,power_law_1.2,0.13025920391082763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,power_law_1.2,0.1385151982307434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,power_law_1.2,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,power_law_1.2,0.1508031964302063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,power_law_1.2,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,power_law_1.2,0.1694591999053955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.18738559484481812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.24759678840637206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.38528640270233155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.5134528160095215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.6389823913574219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,power_law_1.2,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,power_law_1.2,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,power_law_1.2,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,power_law_1.2,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,power_law_1.2,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,power_law_1.2,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,power_law_1.2,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,power_law_1.2,0.10108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,power_law_1.2,0.13289599418640136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.09419519901275634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.14979200363159179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.19591679573059081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.1865407943725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,power_law_1.2,0.14343680143356324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.1937343955039978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.19467519521713256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.1770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.17466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.1720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.179692804813385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.18606079816818238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.17274880409240723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.1774783968925476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.1917631983757019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.19151359796524048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.22549760341644287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.23074560165405272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.27735679149627684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.26998400688171387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.33140480518341064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.3609407901763916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,power_law_1.01,0.46883201599121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,power_law_1.01,0.546451187133789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,power_law_1.01,0.7759359836578369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,power_law_1.01,0.9665535926818848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,power_law_1.2,0.20186879634857177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,power_law_1.2,0.24228479862213134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,power_law_1.2,0.34189438819885254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,power_law_1.2,0.8946687698364257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,power_law_1.2,0.44062080383300783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,power_law_1.2,0.6377088069915772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.1828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,power_law_1.2,0.8205439567565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,power_law_1.01,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,power_law_1.01,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,power_law_1.01,0.04821760058403015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,power_law_1.01,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,power_law_1.01,0.10264320373535156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,power_law_1.01,0.12847360372543334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,power_law_1.01,0.17482240200042726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,power_law_1.01,0.2240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,power_law_1.01,0.31057279109954833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,power_law_1.01,0.39519999027252195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,balanced,0.050240000089009605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,balanced,0.07779199878374736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,balanced,0.10951466361681621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,balanced,0.16750933726628622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,balanced,0.17010666926701865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,balanced,0.1734079917271932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,balanced,0.1763520042101542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,balanced,0.17801066239674887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,balanced,0.18256000677744547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,balanced,0.18348799149195352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,balanced,0.1864853302637736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,balanced,0.18478399515151978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,balanced,0.1902559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,balanced,0.19593065977096558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,balanced,0.2098133365313212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,balanced,0.23365867137908936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,balanced,0.25066665808359784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,balanced,0.288975993792216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,balanced,0.32249067227045697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,balanced,0.43862398465474445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,balanced,0.48661335309346515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,balanced,0.683679978052775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,balanced,0.8286879857381185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,balanced,1.1925280094146729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,balanced,1.531925360361735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.06810240149497986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.1264575958251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.1265663981437683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.13233920335769653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.13422720432281493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.14001280069351196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.14156160354614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.13969279527664186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.14438400268554688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.14612480401992797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.15152000188827514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.16307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.16847360134124756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.1829632043838501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.2074496030807495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.22344961166381835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.26801280975341796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.31112959384918215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,power_law_1.01,0.4083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,power_law_1.01,0.4990719795227051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,power_law_1.01,0.6794432163238525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,power_law_1.01,0.872704029083252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,power_law_1.2,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,power_law_1.2,0.02720000147819519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,power_law_1.2,0.027961599826812743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,power_law_1.2,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,power_law_1.2,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,power_law_1.2,0.03184640109539032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,power_law_1.2,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,power_law_1.2,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,power_law_1.2,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,power_law_1.2,0.13589760065078735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,power_law_1.2,0.15839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,power_law_1.2,0.22924160957336426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,power_law_1.01,0.06721280217170715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,power_law_1.01,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,power_law_1.01,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,power_law_1.01,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,power_law_1.01,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,power_law_1.01,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,power_law_1.01,0.09256319999694824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,power_law_1.01,0.09254400134086609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,power_law_1.01,0.09778559803962708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,power_law_1.01,0.10332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,power_law_1.01,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,power_law_1.01,0.12187520265579224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,power_law_1.01,0.12551679611206054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,power_law_1.01,0.1487615942955017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,power_law_1.01,0.17303680181503295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,power_law_1.01,0.20506880283355713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,power_law_1.01,0.2955199956893921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,power_law_1.01,1.2479231834411622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,power_law_1.01,0.3556544065475464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,power_law_1.2,0.324947190284729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,power_law_1.01,0.38023040294647215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,power_law_1.01,0.6313216209411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,power_law_1.01,0.7326720237731934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,power_law_1.01,1.6177343368530273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,power_law_1.2,0.45915517807006834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,power_law_1.01,1.112723159790039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,power_law_1.2,0.6372992038726807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,power_law_1.01,1.4581184387207031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,power_law_1.01,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,power_law_1.01,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,power_law_1.01,0.10370559692382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,power_law_1.01,0.10417280197143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,power_law_1.01,0.12697600126266478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,power_law_1.01,0.14985599517822265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,power_law_1.01,0.1462656021118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,power_law_1.01,0.17549439668655395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,power_law_1.01,0.18239359855651854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,power_law_1.01,0.1755776047706604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,power_law_1.01,0.17417600154876708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,power_law_1.01,0.18702720403671264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,power_law_1.01,0.19436800479888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,power_law_1.01,0.2036288022994995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,power_law_1.01,0.20641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,power_law_1.01,0.20965120792388917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,power_law_1.01,0.22868480682373046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,power_law_1.01,0.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,power_law_1.01,0.27850239276885985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,power_law_1.01,0.31415040493011476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,power_law_1.01,0.3505664110183716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,power_law_1.01,0.45656957626342776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,power_law_1.01,0.5491968154907226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,power_law_1.01,0.7380864143371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,power_law_1.01,0.8787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,power_law_1.01,1.2172608375549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,power_law_1.01,1.5700927734375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,power_law_1.01,0.0296640008687973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,power_law_1.01,0.029785600304603577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,power_law_1.01,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,power_law_1.01,0.04275839924812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,power_law_1.01,0.049932798743247984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,power_law_1.01,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,power_law_1.01,0.10001280307769775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,power_law_1.01,0.11841919422149658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,power_law_1.01,0.15387519598007202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,power_law_1.01,0.1966271996498108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,power_law_1.01,0.2173504114151001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,power_law_1.01,0.32215681076049807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,power_law_1.01,0.39144320487976075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,power_law_1.01,0.5224319934844971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,power_law_1.01,0.7846848011016846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,power_law_1.01,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,power_law_1.01,0.09579520225524903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,power_law_1.01,0.14598400592803956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,power_law_1.01,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,power_law_1.01,0.1939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,power_law_1.01,0.22316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,power_law_1.01,0.24804480075836183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,power_law_1.01,0.314630389213562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,power_law_1.01,0.2695103883743286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,power_law_1.01,0.2667455911636353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,power_law_1.01,0.3050623893737793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,power_law_1.01,0.3034048080444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,power_law_1.01,0.3119935989379883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,power_law_1.01,0.3201535940170288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,power_law_1.01,0.3404288053512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,power_law_1.01,0.3404927968978882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,power_law_1.01,0.3425663948059082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,power_law_1.01,0.38181118965148925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,power_law_1.01,0.4053184032440186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,power_law_1.01,0.46550397872924804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,power_law_1.01,0.49640960693359376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,power_law_1.01,0.6133503913879395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,power_law_1.01,0.7138432025909424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,power_law_1.01,0.9126399993896485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,power_law_1.01,1.0756671905517579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,power_law_1.01,1.4953344345092774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,power_law_1.01,1.8036159515380858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,balanced,0.05909866591294607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,balanced,0.06051200131575266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,balanced,0.05936000247796377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,balanced,0.08916266759236653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,balanced,0.10236799716949463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,balanced,0.10504532853762309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,balanced,0.10263466835021973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,balanced,0.10552533467610677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,balanced,0.10565333565076192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,balanced,0.10738133390744527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,balanced,0.10949866970380147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,balanced,0.10897599657376607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,balanced,0.11756267150243123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,balanced,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,balanced,0.12621333201726279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,balanced,0.13012799620628357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,balanced,0.14029332995414734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,balanced,0.1584053337574005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,balanced,0.1707520087560018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,balanced,0.21010667085647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,balanced,0.23695466915766397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,balanced,0.298362672328949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,balanced,0.38142398993174237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,balanced,0.510042667388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,balanced,0.6669600009918213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.09649919867515563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.11830400228500366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.11877119541168213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.12027519941329956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.12540800571441652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.12677119970321654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.1392832040786743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.14184319972991943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.1472383975982666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.15030399560928345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.1654080033302307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,power_law_1.01,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.1792448043823242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,power_law_1.01,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.21263360977172852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,power_law_1.01,0.10553599596023559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,power_law_1.01,0.1091647982597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,power_law_1.01,0.13055360317230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.22213120460510255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,power_law_1.01,0.11778559684753417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,power_law_1.01,0.12885760068893432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,power_law_1.01,0.12892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,power_law_1.01,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,power_law_1.01,0.13829760551452636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.23817598819732666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,power_law_1.01,0.14216320514678954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,power_law_1.01,0.15569920539855958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,power_law_1.01,0.16060800552368165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,power_law_1.01,0.1802623987197876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,power_law_1.01,0.20743041038513182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,power_law_1.01,0.20616960525512695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,power_law_1.01,0.25499520301818845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,power_law_1.01,0.29345920085906985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,power_law_1.01,0.3748800039291382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,power_law_1.01,0.48412160873413085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.2957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,power_law_1.01,0.6329023838043213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,power_law_1.01,0.81080322265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,power_law_1.01,1.0980799674987793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,power_law_1.01,1.3931584358215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,power_law_1.2,0.33966081142425536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,power_law_1.2,0.4584383964538574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,power_law_1.2,0.5949376106262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,power_law_1.2,0.8716032028198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,power_law_1.2,1.1694144248962401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,power_law_1.2,1.5670144081115722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,power_law_1.2,2.116204833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,power_law_1.2,2.5497600555419924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.10240639448165893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.12981120347976685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.14034559726715087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.1614143967628479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.1751423954963684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.17998080253601073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.18173439502716066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.18681600093841552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.1710911989212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.1840831995010376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.1922816038131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.19997440576553344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.21191039085388183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.21741440296173095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.233459210395813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.25798399448394777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.3008064031600952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.32230401039123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.39799039363861083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.47582077980041504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,power_law_1.01,0.6082367897033691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,power_law_1.01,0.7323008060455323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,power_law_1.01,1.033670425415039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,power_law_1.01,1.2821696281433106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,power_law_1.2,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,power_law_1.2,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,power_law_1.2,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,power_law_1.2,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,power_law_1.2,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,power_law_1.2,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,power_law_1.2,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,power_law_1.2,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,power_law_1.2,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.2,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.2,0.14120320081710816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.2,0.16232320070266723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.2,0.24129281044006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.2,0.3036288022994995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.2,0.47386879920959474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.2,0.6761536121368408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.2,0.9893055915832519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.2,1.4556991577148437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,balanced,0.04461333155632019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,balanced,0.05288533369700114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,balanced,0.07768533130486806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,balanced,0.07957333326339722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,balanced,0.0786293347676595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,balanced,0.08141866823037465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,balanced,0.08062399923801422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,balanced,0.08271466692288716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,balanced,0.0858133335908254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,balanced,0.08829333384831746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,balanced,0.09330133597056071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,balanced,0.0990613301595052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,balanced,0.10295466581980388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,balanced,0.1162506639957428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,balanced,0.12470400333404541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,balanced,0.1623199979464213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,balanced,0.18698134024937949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,balanced,0.254314661026001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,balanced,0.29234133164087933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,balanced,0.4217280149459839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,balanced,0.5254026651382446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,balanced,0.7460853258768717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,balanced,0.9730292956034342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,power_law_1.2,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,power_law_1.2,0.04782080054283142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,power_law_1.2,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,power_law_1.2,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,power_law_1.2,0.09337599873542786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,power_law_1.2,0.10859520435333252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,power_law_1.2,0.12515840530395508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,power_law_1.2,0.17418880462646485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,power_law_1.2,0.23183999061584473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,power_law_1.2,0.3391616106033325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,power_law_1.2,0.42705278396606444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,power_law_1.2,0.6662911891937255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,power_law_1.2,0.9190527915954589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,power_law_1.01,0.028678399324417115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,power_law_1.01,0.02707200050354004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,power_law_1.01,0.027859199047088622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,power_law_1.01,0.02736000120639801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,power_law_1.01,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,power_law_1.01,0.02852480113506317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,power_law_1.01,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,power_law_1.01,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,power_law_1.01,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,power_law_1.01,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,power_law_1.01,0.08971520066261292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,power_law_1.01,0.12039680480957031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,power_law_1.01,0.18547199964523314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,power_law_1.01,0.23605759143829347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.051526397466659546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.0821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.12361600399017333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.12991360425949097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.1323583960533142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.1342463970184326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.1364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.1425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.14526720046997071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.14643839597702027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.1648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.17055360078811646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.1877120018005371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.22459518909454346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.25484800338745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.3190975904464722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3773952007293701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.519155216217041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6436416149139405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9033023834228515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.1531007766723633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.6676095962524413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.1535295486450194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,power_law_1.2,0.0745024025440216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,power_law_1.2,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,power_law_1.2,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,power_law_1.2,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,power_law_1.2,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,power_law_1.2,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,power_law_1.2,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.2,0.0917248010635376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.2,0.10933120250701904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.2,0.12335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.2,0.15217920541763305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.2,0.1791551947593689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.2,0.24298880100250245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.2,0.30046079158782957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.2,0.4255104064941406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.2,0.5600255966186524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,power_law_1.01,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,power_law_1.01,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,power_law_1.01,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,power_law_1.01,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,power_law_1.01,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,power_law_1.01,0.05112959742546082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,power_law_1.01,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,power_law_1.01,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,power_law_1.01,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,power_law_1.01,0.14704639911651612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,power_law_1.01,0.19213440418243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,power_law_1.01,0.22902400493621827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,power_law_1.01,0.303769588470459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,power_law_1.01,0.38490240573883056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,power_law_1.01,0.5752960205078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,power_law_1.01,0.7502016067504883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.06743040084838867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.1134719967842102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.13184640407562256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.18157440423965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.19005440473556517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.1947711944580078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.19901440143585206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.20245120525360108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.20758399963378907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.21293439865112304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.21872639656066895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.22984960079193115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.24063360691070557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.24763519763946534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.26981120109558104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.3153023958206177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.3543040037155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.44142718315124513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.5160831928253173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.6894464015960693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.8425344467163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.1649087905883788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.5647744178771972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.1830400466918944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,2.7889984130859373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,balanced,0.035258665680885315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,balanced,0.03646933287382126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,balanced,0.03793599953254064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,balanced,0.05843733251094818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,balanced,0.09760000308354695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,balanced,0.11346667011578877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,balanced,0.1479093333085378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,balanced,0.17965332667032877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,balanced,0.2542666594187419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,balanced,0.3278026580810547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,power_law_1.2,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,power_law_1.2,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,power_law_1.2,0.0460671991109848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,power_law_1.2,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,power_law_1.2,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,power_law_1.2,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,power_law_1.2,0.07936000227928161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,power_law_1.2,0.08385279774665833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,power_law_1.2,0.10053759813308716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,power_law_1.2,0.11656320095062256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,power_law_1.2,0.14954240322113038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,power_law_1.2,0.16813440322875978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,power_law_1.2,0.23361918926239014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,power_law_1.2,0.29806079864501955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,power_law_1.2,0.4145088195800781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,power_law_1.2,0.555731201171875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,balanced,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,balanced,0.05763733386993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,balanced,0.05964266757170359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,balanced,0.06604800124963124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,balanced,0.0684799998998642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,balanced,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,balanced,0.08384533723195393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,balanced,0.10710933804512024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,balanced,0.13970133662223816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,balanced,0.17495999733606973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,balanced,0.20189867417017618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,balanced,0.27935999631881714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,balanced,0.3493386507034302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,balanced,0.5008906523386637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,balanced,0.639573335647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,power_law_1.2,0.8045824050903321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,power_law_1.2,1.1310144424438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,power_law_1.2,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,power_law_1.2,0.06384639739990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,power_law_1.2,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,power_law_1.2,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,power_law_1.2,0.08693119883537292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,power_law_1.2,0.09831039905548096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,power_law_1.2,0.11373440027236939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,power_law_1.2,0.13395839929580688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,power_law_1.2,0.15829119682312012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,power_law_1.2,0.19007999897003175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,power_law_1.2,0.2630399942398071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,power_law_1.2,0.3535104036331177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,power_law_1.2,0.5555136203765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,balanced,0.07164266705513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,balanced,0.12174399693806966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,balanced,0.18428266048431396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,balanced,0.30718932549158734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,balanced,0.5484586556752523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,balanced,0.7903839747111002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,balanced,0.7843680381774902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,balanced,0.7864267031351725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,balanced,0.7906080087025961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,balanced,0.7939306894938151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,balanced,0.8017600377400717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,balanced,0.80076797803243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,balanced,0.8073226610819498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,balanced,0.821669340133667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,balanced,0.8252800305684408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,balanced,0.83897598584493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,balanced,0.8641760349273682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,balanced,0.884218692779541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,balanced,0.9372533162434896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,balanced,0.9797333081563314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,balanced,1.09933336575826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.2009173234303792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.4017066955566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,balanced,1.708229382832845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,balanced,2.0953920682271323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,balanced,2.862863858540853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,power_law_1.2,0.6992576122283936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,power_law_1.2,0.8742655754089356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.1623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.192249596118927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.08724480271339416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.27675518989562986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.3845695972442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.08853759765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.49793281555175783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.092985600233078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.8615232467651367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.9233663558959961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.08979840278625488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.9673088073730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.09304320216178893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.09376639723777772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.9901951789855957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.09825279712677001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,1.0194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.09722239971160888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,1.06046724319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.11575039625167846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,1.090022373199463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.1332927942276001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,1.101529598236084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.19441280364990235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.21905279159545898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,1.1569472312927247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.3009727954864502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.3769023895263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,1.1945152282714844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,power_law_1.01,0.5400512218475342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,1.2285823822021484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,power_law_1.01,0.6544576168060303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,1.2802047729492188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.4348544120788573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.5591744422912597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,1.8602048873901367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,2.0443519592285155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,2.587552070617676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,3.035615921020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,4.018291091918945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,4.978559875488282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.2,6.998124694824218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.2,8.994534301757813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,power_law_1.01,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,power_law_1.01,0.10455679893493652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,power_law_1.01,0.15208319425582886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,power_law_1.01,0.24639360904693602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,power_law_1.01,0.4044928073883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,power_law_1.01,0.5547711849212646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,power_law_1.01,0.6480703830718995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,power_law_1.01,0.8307711601257324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,power_law_1.01,0.874835205078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,power_law_1.01,0.8774720191955566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,power_law_1.01,0.950432014465332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,power_law_1.01,0.9862015724182129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,power_law_1.01,1.0126208305358886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,power_law_1.01,1.0357760429382323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,power_law_1.01,1.0659711837768555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,power_law_1.01,1.0821056365966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,power_law_1.01,1.143558406829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,power_law_1.01,1.2275903701782227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,power_law_1.01,1.2918208122253418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,power_law_1.01,1.4448512077331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,power_law_1.01,1.5499648094177245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,power_law_1.01,1.8472639083862306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,power_law_1.01,1.939206314086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,power_law_1.01,2.3862720489501954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,power_law_1.01,2.9693439483642576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,power_law_1.01,3.836851119995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,power_law_1.01,4.728684616088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.01,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.01,0.029529601335525513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.01,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.01,0.02951039969921112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.01,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.01,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.01,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.01,0.031891199946403506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.01,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.01,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.01,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.01,0.047731199860572816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.01,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.01,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.01,0.1353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.01,0.1568063974380493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.16195839643478394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.23345279693603516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.20849919319152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.21553919315338135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.2668735980987549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.2740288019180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.2878976106643677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.29217920303344724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.28263039588928224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.2938175916671753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.30215039253234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.30542080402374266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.31975679397583007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.3223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.33148798942565916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.3351360082626343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.3613312005996704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.422105598449707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.4266176223754883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.5256127834320068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,power_law_1.01,0.6121088027954101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,power_law_1.01,0.713484811782837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,power_law_1.01,0.6638527870178222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,power_law_1.01,0.8494463920593261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,power_law_1.01,1.0865407943725587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,power_law_1.01,1.503980827331543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,power_law_1.01,1.721228790283203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.11751680374145508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.15870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.2092992067337036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.3081023931503296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.33495678901672366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.336844801902771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.3500864028930664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.3499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.36440320014953614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.37150719165802004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.3736000061035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.3822144031524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.39866878986358645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.40682239532470704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.4172544002532959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.4563007831573486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.48917760848999026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.570688009262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.5858431816101074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,power_law_1.01,0.7210559844970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,power_law_1.01,0.843507194519043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,power_law_1.01,1.1150976181030274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,power_law_1.01,1.3887935638427735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,power_law_1.01,1.8941247940063477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,power_law_1.01,2.4010751724243162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.01,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.01,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.01,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.01,0.029126399755477907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.01,0.02956160008907318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.01,0.029516801238059998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.01,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.01,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.01,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.01,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.01,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.01,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.01,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.01,0.12423039674758911
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.01,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,power_law_1.01,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,power_law_1.01,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,power_law_1.01,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,power_law_1.01,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,power_law_1.01,0.05968639850616455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.01,0.08616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.01,0.09816319942474365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.01,0.13265279531478882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.01,0.1635648012161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.01,0.2191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.01,0.3248447895050049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.01,0.4857344150543213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.01,0.6238656044006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,power_law_1.01,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,power_law_1.01,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,power_law_1.01,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,power_law_1.01,0.10454399585723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,power_law_1.01,0.11137280464172364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,power_law_1.01,0.12191359996795655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,power_law_1.01,0.1314303994178772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,power_law_1.01,0.13198720216751098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,power_law_1.01,0.13214720487594606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,power_law_1.01,0.1424512028694153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,power_law_1.01,0.1538624048233032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,power_law_1.01,0.16545920372009276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,power_law_1.01,0.17555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,power_law_1.01,0.19814399480819703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,power_law_1.01,0.23339519500732422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,power_law_1.01,0.23947520256042482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,power_law_1.01,0.2939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,power_law_1.01,0.34491519927978515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,power_law_1.01,0.461740779876709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,power_law_1.01,0.5865407943725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,power_law_1.01,0.779097604751587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,power_law_1.01,0.989145565032959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,power_law_1.01,1.443391990661621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,power_law_1.01,1.8130559921264648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,power_law_1.2,0.10500479936599731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,power_law_1.2,0.11334400177001953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,power_law_1.2,0.13999999761581422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,power_law_1.2,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,power_law_1.2,0.16297600269317628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,power_law_1.2,0.16082559823989867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,power_law_1.2,0.16636159420013427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,power_law_1.2,0.17534079551696777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,power_law_1.2,0.17406079769134522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,power_law_1.2,0.16227200031280517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,power_law_1.2,0.17630079984664918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,power_law_1.2,0.1827712059020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,power_law_1.2,0.1879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,power_law_1.2,0.20302720069885255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,power_law_1.2,0.2114176034927368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,power_law_1.2,0.21742720603942872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,power_law_1.2,0.22551679611206055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,power_law_1.2,0.2698048114776611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,power_law_1.2,0.27852160930633546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,power_law_1.2,0.345196795463562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,power_law_1.2,0.3949631929397583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,power_law_1.2,0.5134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,power_law_1.2,0.532038402557373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,power_law_1.2,0.7318079948425293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,power_law_1.2,0.9539072036743164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,power_law_1.2,1.2273599624633789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,power_law_1.2,1.6142208099365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,1,balanced,0.18561599651972452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,1,balanced,0.1913706660270691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,1,balanced,0.2145813306172689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,1,balanced,0.27511467536290485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,1,balanced,0.42290135224660236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,1,balanced,0.7160800298055013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,1,balanced,0.720911979675293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,1,balanced,0.724554697672526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,1,balanced,0.7251253128051758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,1,balanced,0.7282079855600992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,1,balanced,0.730026642481486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,1,balanced,0.7369493643442789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,1,balanced,0.741376002629598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,1,balanced,0.7492160002390543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,1,balanced,0.744767983754476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,1,balanced,0.7542133331298828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,1,balanced,0.7625226974487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,1,balanced,0.7927626768747965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,1,balanced,0.8165919780731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,1,balanced,0.876197338104248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,1,balanced,0.9415466785430908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,1,balanced,1.0354453722635906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,1,balanced,1.1527893543243408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,1,balanced,1.5972960789998372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,1,balanced,1.8399359385172527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,1,balanced,2.5666826566060386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,1,balanced,3.3825012842814126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,power_law_1.01,0.15353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,power_law_1.01,0.24567039012908937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,power_law_1.01,0.43683838844299316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,power_law_1.01,0.6273983955383301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,power_law_1.01,0.8447551727294922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,power_law_1.01,1.1630975723266601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,power_law_1.01,1.2241344451904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,power_law_1.01,1.2750656127929687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,power_law_1.01,1.3103487968444825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,power_law_1.01,1.3607359886169434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,power_law_1.01,1.3920319557189942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,power_law_1.01,1.4410688400268554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,power_law_1.01,1.4509440422058106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,power_law_1.01,1.5150912284851075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,power_law_1.01,1.5318976402282716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,power_law_1.01,1.5619520187377929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,power_law_1.01,1.6496768951416017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,power_law_1.01,1.7612863540649415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,power_law_1.01,1.9413440704345704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,power_law_1.01,1.9645631790161133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,power_law_1.01,2.3274879455566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,power_law_1.01,2.582809638977051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,power_law_1.01,3.141868782043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,power_law_1.01,3.6009281158447264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,power_law_1.01,4.733126449584961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,power_law_1.01,5.7777153015136715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,balanced,0.04538666705290476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,balanced,0.07600000003973643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,balanced,0.10656533638636272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,balanced,0.11598933736483256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,balanced,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,balanced,0.11909332871437073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,balanced,0.11955733100573222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,balanced,0.11961066722869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,balanced,0.12125333150227864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,balanced,0.12152000268300374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,balanced,0.12284266948699951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,balanced,0.12627200285593668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,balanced,0.13083199659983316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,balanced,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,balanced,0.13726933797200522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,balanced,0.1534293293952942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,balanced,0.1579253375530243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,balanced,0.18650666872660318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,balanced,0.19801066319147745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,balanced,0.2505226731300354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,balanced,0.2871840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,balanced,0.40573867162068683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,balanced,0.4706079959869385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,balanced,0.657589316368103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,balanced,0.8334293365478516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.06347519755363465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.11754239797592163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.14495999813079835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.14952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.1492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.16013439893722534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.15920000076293944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.16383999586105347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.17533440589904786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.17717119455337524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.19393279552459716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.2199552059173584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.2382591962814331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.28666880130767824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.3420480012893677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.44037761688232424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.574015998840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.7538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,power_law_1.01,0.910912036895752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.2712063789367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,power_law_1.01,1.620102310180664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,balanced,0.07771733403205872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,balanced,0.11370133360226949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,balanced,0.17315733432769775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,balanced,0.28225066264470416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,balanced,0.3851199944814046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,balanced,0.38768001397450763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,balanced,0.3903573354085286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,balanced,0.39157867431640625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,balanced,0.3975040117899577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,balanced,0.40456533432006836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,balanced,0.4063733418782552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,balanced,0.4205813407897949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,balanced,0.43036798636118573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,balanced,0.44113067785898846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,balanced,0.46191998322804767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,balanced,0.5139626661936442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,balanced,0.5598826805750529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,balanced,0.6422666708628336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,balanced,0.7155199845631918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,balanced,0.8834773699442545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,balanced,1.071552038192749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,balanced,1.3779360453287761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,balanced,1.8117599487304688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,balanced,2.449589411417643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,balanced,3.285898526509603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,power_law_1.01,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,power_law_1.01,0.052076798677444455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,power_law_1.01,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,power_law_1.01,0.0782144010066986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,power_law_1.01,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,power_law_1.01,0.08426240086555481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,power_law_1.01,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,power_law_1.01,0.08702080249786377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,power_law_1.01,0.0887935996055603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,power_law_1.01,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,power_law_1.01,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,power_law_1.01,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,power_law_1.01,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,power_law_1.01,0.10769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,power_law_1.01,0.12176640033721924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,power_law_1.01,0.14396799802780152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,power_law_1.01,0.17132799625396727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,power_law_1.01,0.22577280998229982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,power_law_1.01,0.288428807258606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,power_law_1.01,0.4194047927856445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,power_law_1.01,0.5257215976715088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,power_law_1.01,0.7349696159362793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,power_law_1.01,0.9965888023376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.08202880024909973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.10883840322494506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.11726720333099365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.1195072054862976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.12106239795684814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.11900160312652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.12254719734191895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.13946880102157594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.1473855972290039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.14703359603881835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.16513919830322266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.18951040506362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.22895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.25892479419708253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.31303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.38904318809509275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.5068672180175782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,power_law_1.2,0.6617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,power_law_1.2,0.90830078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,power_law_1.2,1.1260479927062987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.07007359862327575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.07857919931411743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.08352000117301941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.09075199961662292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.09844480156898498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.11199359893798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.13285759687423707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1472000002861023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.18644479513168336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.21989760398864747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.30169599056243895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.35960960388183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.5143487930297852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.6824192047119141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.08936960101127625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.11921279430389405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.16767359972000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2227328062057495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.2663615942001343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.3687360048294067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,balanced,0.07404266794522603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,balanced,0.10123733679453532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,balanced,0.15081066886583963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,balanced,0.258517324924469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,balanced,0.4699360132217407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,balanced,0.5765386819839478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,balanced,0.5805600086847941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,balanced,0.5820266803105673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,balanced,0.5866186618804932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,balanced,0.5887893438339233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,balanced,0.5961386760075887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,balanced,0.6010239919026693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,balanced,0.6050453186035156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,balanced,0.6136213143666586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,balanced,0.6303253173828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,balanced,0.635322650273641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,balanced,0.654149333635966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,balanced,0.75055464108785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,balanced,0.7258986632029215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,balanced,0.889520009358724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,balanced,0.8777013619740804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,balanced,1.3103946844736736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,balanced,1.215488036473592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,balanced,1.7807572682698567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,balanced,2.034245332082113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,balanced,3.314885457356771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,balanced,3.75763734181722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.502348804473877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,power_law_1.01,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,power_law_1.01,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,power_law_1.01,0.04177280068397522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,power_law_1.01,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,power_law_1.01,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,power_law_1.01,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,power_law_1.01,0.0555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,power_law_1.01,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.01,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.01,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.01,0.14944000244140626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.01,0.19366400241851806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.01,0.2589375972747803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.01,0.3757823944091797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.01,0.6009920120239258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.01,0.7907008171081543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.04999679923057556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.07004160284996033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.10490239858627319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.13753600120544435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.16476160287857056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.23059840202331544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.29087998867034914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.41852798461914065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.5444479942321777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,balanced,0.057706668972969055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,balanced,0.07018133501211803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,balanced,0.07217066486676534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,balanced,0.07270933190981548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,balanced,0.07878933350245158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,balanced,0.08583999673525493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,balanced,0.08673066894213359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,balanced,0.09282132983207703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,balanced,0.10521066188812256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,balanced,0.11547733346621196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,balanced,0.14057599504788718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,balanced,0.16451733311017355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,balanced,0.21501866976420084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,balanced,0.2299306591351827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,balanced,0.2936426599820455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,balanced,0.35711999734242755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,balanced,0.4997440179189046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,balanced,0.6179306507110596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,balanced,0.06477866570154826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,balanced,0.09340266386667888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,balanced,0.13900267084439596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,balanced,0.21011734008789062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,balanced,0.21332800388336182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,balanced,0.2181333303451538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,balanced,0.21922133366266885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,balanced,0.22202134132385254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,balanced,0.226800004641215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,balanced,0.2319200038909912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,balanced,0.23388799031575522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,balanced,0.24316267172495523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,balanced,0.25038933753967285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,balanced,0.2592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,balanced,0.2773653268814087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,balanced,0.3171520034472148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,balanced,0.34777601559956867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,balanced,0.42348265647888184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,balanced,0.49249064922332764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,balanced,0.6895039876302084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,balanced,0.802010695139567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,balanced,1.1809386412302654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,balanced,1.4659306208292644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,balanced,2.139397303263346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,balanced,2.7837546666463218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,power_law_1.01,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,power_law_1.01,0.09843840003013611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,power_law_1.01,0.11467519998550416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,power_law_1.01,0.11928319931030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,power_law_1.01,0.12362879514694214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,power_law_1.01,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,power_law_1.01,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,power_law_1.01,0.13530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,power_law_1.01,0.13994239568710326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,power_law_1.01,0.14650239944458007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,power_law_1.01,0.15342719554901124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,power_law_1.01,0.16677119731903076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,power_law_1.01,0.20266239643096923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,power_law_1.01,0.2207103967666626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,power_law_1.01,0.2748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,power_law_1.01,0.3118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,power_law_1.01,0.4112383842468262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,power_law_1.01,0.5130047798156738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,power_law_1.01,0.6939712047576905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,power_law_1.01,0.8755583763122559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,power_law_1.01,1.244480037689209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,power_law_1.01,1.614348793029785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,balanced,0.043247997760772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,balanced,0.04884799818197886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,balanced,0.05269333223501841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,balanced,0.07346133391062419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,balanced,0.07612266639868419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,balanced,0.07672533392906189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,balanced,0.07655466596285503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,balanced,0.0794239987929662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,balanced,0.0888213316599528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,balanced,0.08969066540400188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,balanced,0.09371733665466309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,balanced,0.1043893297513326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,balanced,0.11028800408045451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,balanced,0.14431466658910116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,balanced,0.16160000363985697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,balanced,0.22215465704600015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,balanced,0.2469386657079061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,balanced,0.36022400856018066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,balanced,0.4300533135732015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,balanced,0.5823946793874105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,balanced,0.7501227060953776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,balanced,0.044794668753941856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.2,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.2,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.2,0.05421440005302429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.2,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.2,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,balanced,0.04958933095137278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.2,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.2,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.2,0.06853119730949402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.2,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.2,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.2,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.2,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.2,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.2,0.11575039625167846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.2,0.14110080003738404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.2,0.16670080423355102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.2,0.21984639167785644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.2,0.23095040321350097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.2,0.30658559799194335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.2,0.4202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.2,0.5839104175567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.2,0.7432127952575683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,balanced,0.04619733492533366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,balanced,0.06446933249632518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,balanced,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,balanced,0.0768693337837855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,balanced,0.10136000315348308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,balanced,0.12301333745320638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,balanced,0.15546133120854697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,balanced,0.18685332934061685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,balanced,0.25336533784866333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,balanced,0.3140373428662618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,balanced,0.44387733936309814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,balanced,0.5704053243001302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,balanced,0.052842666705449425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,balanced,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,balanced,0.07282666862010956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,balanced,0.08288533488909404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,balanced,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,balanced,0.11372266213099162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,balanced,0.12970133622487387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,balanced,0.17061332861582437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,balanced,0.21144000689188638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,balanced,0.29496532678604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,balanced,0.37356265385945636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,balanced,0.04638933142026266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,balanced,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,balanced,0.08867733677228291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,balanced,0.13316266735394797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,balanced,0.13711466391881308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,balanced,0.13180800278981528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,balanced,0.13408000270525613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,balanced,0.13424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,balanced,0.1351093351840973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,balanced,0.13843199610710144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,balanced,0.14032533764839172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,balanced,0.14223466316858926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,balanced,0.1523146629333496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,balanced,0.15263467033704123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,balanced,0.1614346702893575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,balanced,0.18062400817871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,balanced,0.19615467389424643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,balanced,0.23533332347869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,balanced,0.26898666222890216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,balanced,0.37812801202138263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,balanced,0.42981334527333576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,balanced,0.6249333222707113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,balanced,0.763744036356608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,balanced,1.1163679758707683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,balanced,1.4045012791951497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,balanced,0.02533866713444392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,balanced,0.02754133443037669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,balanced,0.02956799914439519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,balanced,0.06187733511130015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,balanced,0.052469333012898765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,balanced,0.0537120004494985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,balanced,0.053632001082102455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,balanced,0.06448533137639363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,balanced,0.0662773350874583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,balanced,0.07057066758473714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,balanced,0.08210133512814839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,balanced,0.08984532952308655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,balanced,0.09109333157539368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,balanced,0.11181867122650146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,balanced,0.12961600224177042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,balanced,0.16942399740219116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,balanced,0.19986667235692343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,balanced,0.27512532472610474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,balanced,0.3408799966176351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.14595839977264405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.17787519693374634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.21960320472717285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.27562239170074465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.3464960098266602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.11145600080490112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.38646399974823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.36583681106567384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.11041920185089112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.3573632001876831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.3801215887069702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.12098560333251954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.3999295949935913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.13226239681243895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.404633617401123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.39358720779418943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.136518394947052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.4039423942565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.1614143967628479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.4158592224121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.17466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.417574405670166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.2046976089477539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.44049921035766604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.49310078620910647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.48876161575317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.5615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,power_law_1.01,0.5771584033966064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,power_law_1.01,0.6685823917388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.2915008068084717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,power_law_1.01,0.7576320171356201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.36469120979309083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.5148352146148681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,power_law_1.01,0.9288384437561035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.6495488166809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,power_law_1.01,1.1186304092407227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,0.9263744354248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,power_law_1.01,1.4273664474487304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.1568384170532227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,power_law_1.01,1.7265792846679688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,balanced,0.03121600051720937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,balanced,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,balanced,0.03299200038115183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,balanced,0.03294399877389272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,balanced,0.03410666684309641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,balanced,0.05073600014050802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,balanced,0.06814933319886525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,balanced,0.08867200215657552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,balanced,0.09703466296195984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,balanced,0.04628799855709076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,balanced,0.04763199885686239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,balanced,0.04850666721661886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,balanced,0.07523733377456665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,balanced,0.07607466479142506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,balanced,0.0765066643555959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,balanced,0.07817600170771281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,balanced,0.08126399914423625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,balanced,0.0804799993832906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,balanced,0.08340799808502197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,balanced,0.08849599957466125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,balanced,0.09055999914805095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,balanced,0.09325866897900899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,balanced,0.1092746655146281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,balanced,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,balanced,0.1431893308957418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,balanced,0.17853333552678427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,balanced,0.20928533871968588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,balanced,0.27560534079869586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,balanced,0.3385813236236572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,balanced,0.4771360158920288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,balanced,0.6095306475957235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,balanced,0.03527999917666117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,balanced,0.03528533379236857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,balanced,0.04125866790612539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,balanced,0.0825439989566803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,balanced,0.09516266981760661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,balanced,0.12125333150227864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,balanced,0.1476959983507792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,balanced,0.193615992863973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,balanced,0.24498132864634195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,balanced,0.03908266623814901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,balanced,0.04875733455022176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,balanced,0.05133866767088572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,balanced,0.05039466420809428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,balanced,0.055045331517855324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,balanced,0.08422399560610454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,balanced,0.08295466502507527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,balanced,0.10332799951235454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,balanced,0.11586667100588481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,balanced,0.15346666177113852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,balanced,0.17384000619252524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,balanced,0.23762667179107666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,balanced,0.27958933512369794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,balanced,0.07076266904671986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,balanced,0.08586666981379192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,balanced,0.08642666538556416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,balanced,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,balanced,0.08571199576059978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,balanced,0.0855466624101003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,balanced,0.08546666304270427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,balanced,0.0865226686000824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,balanced,0.08648533622423808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,balanced,0.08923199772834778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,balanced,0.09370666742324829
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,balanced,0.09514133135477702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,balanced,0.0976586639881134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,balanced,0.10280533631642659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,balanced,0.10934399565060933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,balanced,0.13195199767748514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,balanced,0.138565331697464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,balanced,0.18553600708643594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,balanced,0.19870400428771973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,balanced,0.2715946634610494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,balanced,0.3145280083020528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,balanced,0.45398934682210285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,balanced,0.5538186629613241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,balanced,0.04398933549722036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,balanced,0.10505599776903789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,balanced,0.10875733693440755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,balanced,0.10894933342933655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,balanced,0.10896000266075134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,balanced,0.11012799541155498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,balanced,0.11467732985814412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,balanced,0.11286933223406474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,balanced,0.114847997824351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,balanced,0.11777066191037495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,balanced,0.11928000052769978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,balanced,0.1269439955552419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,balanced,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,balanced,0.13372266292572021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,balanced,0.1481653352578481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.1630293329556783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.18914665778477988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.21212265888849893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.28782933950424194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.3373546600341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.48152001698811847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.6011626720428467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.8683040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,balanced,1.1313493251800537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.04275839924812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.11786240339279175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.15796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.16750719547271728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,power_law_1.2,0.2885504007339478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,power_law_1.2,0.3666559934616089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,power_law_1.2,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,power_law_1.2,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,power_law_1.2,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,power_law_1.2,0.07863039970397949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,power_law_1.2,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,power_law_1.2,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,power_law_1.2,0.11416959762573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,power_law_1.2,0.12419840097427368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,power_law_1.2,0.1557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,power_law_1.2,0.18810240030288697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,power_law_1.2,0.2465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,power_law_1.2,0.2872960090637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,power_law_1.2,0.39932799339294434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,power_law_1.2,0.5394815921783447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,power_law_1.2,0.7909887790679931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,power_law_1.2,0.9972800254821778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,power_law_1.01,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,power_law_1.01,0.082777601480484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,power_law_1.01,0.10826879739761353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,power_law_1.01,0.11822079420089722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,power_law_1.01,0.11477119922637939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,power_law_1.01,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,power_law_1.01,0.11915520429611207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,power_law_1.01,0.12233599424362182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,power_law_1.01,0.1267583966255188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,power_law_1.01,0.13273600339889527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,power_law_1.01,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,power_law_1.01,0.14379520416259767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,power_law_1.01,0.14739199876785278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,power_law_1.01,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,power_law_1.01,0.184716796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,power_law_1.01,0.21597440242767335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,power_law_1.01,0.2385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,power_law_1.01,0.2967999935150146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,power_law_1.01,0.3553407907485962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,power_law_1.01,0.4872255802154541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,power_law_1.01,0.6257791996002198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,power_law_1.01,0.8494912147521972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,power_law_1.01,1.0905920028686524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,power_law_1.2,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,power_law_1.2,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,power_law_1.2,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,power_law_1.2,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,power_law_1.2,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,power_law_1.2,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,power_law_1.2,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,power_law_1.2,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,power_law_1.2,0.08510720133781433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,power_law_1.2,0.09436799883842469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,power_law_1.2,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,power_law_1.2,0.14207359552383422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,power_law_1.2,0.19285119771957399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,power_law_1.2,0.235532808303833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,power_law_1.2,0.31946239471435545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,power_law_1.2,0.4178880214691162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,power_law_1.2,0.6057663917541504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,power_law_1.2,0.7947391986846923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,balanced,0.06378133098284404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,balanced,0.0660693347454071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,balanced,0.06694933275381725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,balanced,0.06715199848016103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,balanced,0.06811200082302094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,balanced,0.07060799996058147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,balanced,0.07226133346557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,balanced,0.07710933188597362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,balanced,0.07949333389600118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,balanced,0.08339732885360718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,balanced,0.09300266702969869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,balanced,0.09941333532333374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,balanced,0.11321066816647847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,balanced,0.12548266847928366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,balanced,0.16088533401489258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,balanced,0.1882986625035604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,balanced,0.2545173366864522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,balanced,0.3057600061098735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,balanced,0.4315199851989746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,balanced,0.5520960092544556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.08513280153274536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.10999679565429688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.11534719467163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.13320959806442262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.1543295979499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.18600959777832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.22028799057006837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.3096895933151245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.3991487979888916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.5337344169616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,power_law_1.2,0.7127679824829102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,2,balanced,0.28730666637420654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,2,balanced,0.3138773242632548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,2,balanced,0.42578665415445965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,2,balanced,0.6722346941630045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,2,balanced,1.1207466920216878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,2,balanced,1.5414932568868
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,2,balanced,1.545738697052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,2,balanced,1.5456959406534831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,2,balanced,1.550154685974121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,2,balanced,1.554634730021159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,2,balanced,1.5582186381022136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,2,balanced,1.573354721069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,2,balanced,1.571573257446289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,2,balanced,1.585546652475993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,2,balanced,1.586517333984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,2,balanced,1.5964694023132324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,2,balanced,1.6203413009643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,2,balanced,1.6765066782633464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,2,balanced,1.7187946637471516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,2,balanced,1.8338027000427246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,2,balanced,1.9282399813334148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,2,balanced,2.581162611643473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,2,balanced,2.779226620992025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,2,balanced,3.800821304321289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,2,balanced,4.245029449462891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,2,balanced,6.4760691324869795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,2,balanced,8.074000040690104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,balanced,0.05151999990145365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,balanced,0.11778666575749715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,balanced,0.20242132743199667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,balanced,0.37009600798288983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,balanced,0.37161068121592206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,balanced,0.3736426830291748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,balanced,0.37670934200286865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,balanced,0.3773813247680664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,balanced,0.38011733690897626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,balanced,0.38284265995025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,balanced,0.38628800710042316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,balanced,0.38844799995422363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,balanced,0.3951306740442912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,balanced,0.4023733139038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,balanced,0.40882666905721027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,balanced,0.42208532492319745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,balanced,0.44837331771850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,balanced,0.4734239975611369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,balanced,0.5113973220189413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,balanced,0.5614826679229736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,balanced,0.6992053190867106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,balanced,0.7664426962534586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,balanced,1.0413653055826824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,balanced,1.2287413279215496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,balanced,1.7385120391845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,balanced,2.26037867863973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,power_law_1.2,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,power_law_1.2,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,power_law_1.2,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,power_law_1.2,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,power_law_1.2,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,power_law_1.2,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,power_law_1.2,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,power_law_1.2,0.05161600112915039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,power_law_1.2,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,power_law_1.2,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,power_law_1.2,0.12392319440841675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,power_law_1.2,0.1704192042350769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,power_law_1.2,0.21402239799499512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,power_law_1.2,0.3204735994338989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,power_law_1.2,0.40259838104248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,8,balanced,0.08292266726493835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,8,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,8,balanced,0.10410133004188538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,8,balanced,0.1318186620871226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,8,balanced,0.1906613310178121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,8,balanced,0.2996266682942708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,8,balanced,0.4134720166524251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,8,balanced,0.41630931695302326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,8,balanced,0.4133760134379069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,8,balanced,0.4166133403778076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,8,balanced,0.4166933298110962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,8,balanced,0.4176906744639079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,8,balanced,0.42022399107615155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,8,balanced,0.4223039944966634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,8,balanced,0.43145068486531574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,8,balanced,0.4337173302968343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,8,balanced,0.4392640193303426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,8,balanced,0.44947731494903564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,8,balanced,0.4575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,8,balanced,0.48207465807596844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,8,balanced,0.5011680126190186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,8,balanced,0.5550826787948608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,8,balanced,0.6021920045216879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,8,balanced,0.7096959749857584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,8,balanced,0.9174346923828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,8,balanced,1.130090634028117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,8,balanced,1.4857333501180012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,power_law_1.01,0.08513919711112976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,power_law_1.01,0.11148159503936768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,power_law_1.01,0.10682239532470703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,power_law_1.01,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,power_law_1.01,0.11164159774780273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,power_law_1.01,0.09772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,power_law_1.01,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,power_law_1.01,0.10430719852447509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,power_law_1.01,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,power_law_1.01,0.10824960470199585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,power_law_1.01,0.10467840433120727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,power_law_1.01,0.10565760135650634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,power_law_1.01,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,power_law_1.01,0.11915520429611207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,power_law_1.01,0.13207039833068848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,power_law_1.01,0.13883520364761354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,power_law_1.01,0.16332800388336183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,power_law_1.01,0.1679744005203247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,power_law_1.01,0.2063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,power_law_1.01,0.2219327926635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,power_law_1.01,0.29395840167999265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,power_law_1.01,0.3539135932922363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,power_law_1.01,0.49170560836791993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,power_law_1.01,0.6029056072235107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.092985600233078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.12726399898529053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.15255039930343628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.17713919878005982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.19905279874801635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.20088961124420165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.21386239528656006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.21199359893798828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.2243328094482422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.21998720169067382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.23345279693603516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.23447039127349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.24866559505462646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.25642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.26943359375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.2938751935958862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.3049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.3717439889907837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.3915456056594849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.5004032135009766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,power_law_1.2,0.5253568172454834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,power_law_1.2,0.6907904148101807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,power_law_1.2,0.8086463928222656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,power_law_1.2,1.1108480453491212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,power_law_1.2,1.4613951683044433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.15100159645080566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.2427903890609741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.32486400604248045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.3564543962478638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.4817791938781738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.5330944061279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.5028672218322754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.6161471843719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.5895423889160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.5295423984527587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.5157311916351318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.49822077751159666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.5395520210266114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.578547191619873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.5418367862701416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.5582528114318848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.5746816158294678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.6210239887237549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.7096511840820312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,power_law_1.2,0.8527999877929687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,power_law_1.2,0.8552384376525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.0131327629089355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,power_law_1.2,1.07259521484375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,power_law_1.2,1.4298944473266602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,power_law_1.2,1.4379712104797364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,power_law_1.2,1.837459182739258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,power_law_1.2,2.311903953552246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,power_law_1.01,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,power_law_1.01,0.10055040121078491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,power_law_1.01,0.12937599420547485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,power_law_1.01,0.16442240476608277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,power_law_1.01,0.22352640628814696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,power_law_1.01,0.27139840126037595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,power_law_1.01,0.3487423896789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,power_law_1.01,0.35932159423828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,power_law_1.01,0.380729603767395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,power_law_1.01,0.3857151985168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,power_law_1.01,0.3903104066848755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,power_law_1.01,0.408787202835083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,power_law_1.01,0.39946880340576174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,power_law_1.01,0.42381439208984373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,power_law_1.01,0.4407872200012207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,power_law_1.01,0.43628802299499514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,power_law_1.01,0.4656383991241455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,power_law_1.01,0.5160895824432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,power_law_1.01,0.5415872097015381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,power_law_1.01,0.6336448192596436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,power_law_1.01,0.66561279296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,power_law_1.01,0.7980800151824952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,power_law_1.01,0.9180671691894531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,power_law_1.01,1.1942399978637694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,power_law_1.01,1.4537599563598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,power_law_1.01,2.0157119750976564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,power_law_1.01,2.491200065612793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.04200960099697113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.03994880020618439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.05694079995155334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.08831999897956848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.09820160269737244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.10881279706954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.1288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.1474176049232483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.20284159183502198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.25515520572662354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.3703871965408325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.45348482131958007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,balanced,0.04320533573627472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,balanced,0.06157866617043813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,balanced,0.08756800492604573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,balanced,0.08393067121505737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,balanced,0.08539199829101562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,balanced,0.08700266480445862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,balanced,0.08755200107892354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,balanced,0.08757332960764568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,balanced,0.08772800366083781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,balanced,0.09475200374921162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,balanced,0.09382933378219604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,balanced,0.09964799880981445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,balanced,0.10742933551470439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,balanced,0.1136799951394399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,balanced,0.13538666566212973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,balanced,0.142767995595932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,balanced,0.18078400691350302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,balanced,0.1995733380317688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,balanced,0.27669866879781085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,balanced,0.336458683013916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,balanced,0.4801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,balanced,0.6050719817479452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.15431679487228395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.2122431993484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.32371199131011963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.524563217163086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.7166143894195557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.8272895812988281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,1.017612838745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,1.0958271980285645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,1.1224767684936523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,1.1931072235107423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,1.2312000274658204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,1.239129638671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,1.2437248229980469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,1.29934720993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,1.3363264083862305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,1.376959991455078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,1.4800000190734863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.5216511726379394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.6823040008544923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,1.8164991378784179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,2.1625471115112305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,2.3091007232666017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,2.8400127410888674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.01,3.4468223571777346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.01,4.536243057250976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.01,5.6178241729736325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,balanced,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,balanced,0.0558240016301473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,balanced,0.06274133423964183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,balanced,0.06491733094056447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,balanced,0.06446399788061778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,balanced,0.06524266799290974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,balanced,0.06673599779605865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,balanced,0.07262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,balanced,0.07308800021807353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,balanced,0.08301866551240285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,balanced,0.08716799815495808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,balanced,0.10107733805974324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,balanced,0.11384000380833943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,balanced,0.14428266882896423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,balanced,0.17005334297815958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,balanced,0.23443732659022012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,balanced,0.28060267368952435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,balanced,0.392031987508138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,balanced,0.49856531620025635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,power_law_1.2,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,power_law_1.2,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,power_law_1.2,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,power_law_1.2,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,power_law_1.2,0.12076159715652465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,power_law_1.2,0.12040319442749023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,power_law_1.2,0.12596479654312134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,power_law_1.2,0.12799359560012818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,power_law_1.2,0.13452800512313842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,power_law_1.2,0.13813120126724243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,power_law_1.2,0.1419327974319458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,power_law_1.2,0.15096319913864137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,power_law_1.2,0.15573760271072387
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,power_law_1.2,0.17023999691009523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,power_law_1.2,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,power_law_1.2,0.22885119915008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,power_law_1.2,0.2768383979797363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,power_law_1.2,0.31432960033416746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,power_law_1.2,0.4142784118652344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,power_law_1.2,0.5189951896667481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,power_law_1.2,0.695468807220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,power_law_1.2,0.8741312026977539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,power_law_1.2,1.238047981262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,power_law_1.2,1.6121088027954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,balanced,0.05228800078233083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,balanced,0.05775466561317444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,balanced,0.06257066627343495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,balanced,0.06384000182151794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,balanced,0.06638933221499126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,balanced,0.06578666468461354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,balanced,0.06448000172773997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,balanced,0.06831466654936473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,balanced,0.06664533416430156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,balanced,0.07105599840482076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,balanced,0.07658666869004567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,balanced,0.0830506682395935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,balanced,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,balanced,0.1048959990342458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,balanced,0.13078400492668152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,balanced,0.15077867110570273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,balanced,0.18933866421381632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,balanced,0.22809600830078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,balanced,0.30640532573064166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,balanced,0.38394665718078613
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,balanced,0.5377866824467977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,balanced,0.6917866865793864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,balanced,0.04343999922275543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,balanced,0.07017600039641063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,balanced,0.096261332432429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,balanced,0.09745066364606221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,balanced,0.09782399733861287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,balanced,0.09908800323804219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,balanced,0.09877866506576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,balanced,0.10081066687901814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,balanced,0.10096533099810283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,balanced,0.10333333412806193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,balanced,0.10226133465766907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,balanced,0.10920533537864685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,balanced,0.10937600334485371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,balanced,0.1113866666952769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,balanced,0.11913067102432251
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,balanced,0.11988266309102376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,balanced,0.13269866506258646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,balanced,0.14940800269444784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,balanced,0.17279465993245444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,balanced,0.19473600387573242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,balanced,0.2412853240966797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,balanced,0.2956639925638835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,balanced,0.3978613217671712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,balanced,0.4888533353805542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.01,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.01,0.09898239970207215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.01,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.01,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.01,0.06575999855995178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.01,0.0782144010066986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.01,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.01,0.10570240020751953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.01,0.13322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.01,0.1469823956489563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.01,0.18172160387039185
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.01,0.19128960371017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.01,0.24647040367126466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.01,0.35519359111785886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.01,0.48504958152770994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.01,0.5485375881195068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,balanced,0.07863466441631317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,balanced,0.1176533301671346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,balanced,0.12160000205039978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,balanced,0.11573867003122966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,balanced,0.11782399813334148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,balanced,0.11982933680216472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,balanced,0.12155733505884807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,balanced,0.12331733107566833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,balanced,0.127402663230896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,balanced,0.12982933719952902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,balanced,0.13748799761136374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,balanced,0.14282133181889853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,balanced,0.15491200486818948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,balanced,0.1777013341585795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,balanced,0.19463467597961426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,balanced,0.2351306676864624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,balanced,0.27317333221435547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,balanced,0.3970666726430257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,balanced,0.458352009455363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,balanced,0.6747199694315592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,balanced,0.8530986309051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,balanced,1.2450559933980305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,balanced,1.6194879213968914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,balanced,0.039274667700131737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,balanced,0.04552533229192098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,balanced,0.0516533354918162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,balanced,0.0621919979651769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,balanced,0.07130133112271626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,balanced,0.09287466605504353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,balanced,0.10838933785756429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,balanced,0.14088533322016397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,balanced,0.17249600092569986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,balanced,0.070933332045873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,balanced,0.07992533346017201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,balanced,0.08117866516113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,balanced,0.08354666829109192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,balanced,0.08313066760698955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,balanced,0.08276799817879994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,balanced,0.0872213343779246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,balanced,0.0870293378829956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,balanced,0.09290666381518047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,balanced,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,balanced,0.09899733463923137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,balanced,0.11226133505503337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,balanced,0.1179200013478597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,balanced,0.1446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,balanced,0.15709867080052695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,balanced,0.2012373407681783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,balanced,0.2035199999809265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,balanced,0.266810675462087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,balanced,0.29747732480367023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,balanced,0.4071466525395711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,balanced,0.4928640127182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.08157439827919007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.12243839502334594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.13949439525604249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.18313599824905397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.21723520755767822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.3037823915481567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.38565759658813475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.5528639793395996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.7140799999237061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,balanced,0.04349866509437561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,balanced,0.058864002426465355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,balanced,0.06282666822274525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,balanced,0.06884799897670746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,balanced,0.07152533531188965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,balanced,0.0784746656815211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,balanced,0.08390933275222778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,balanced,0.10211199522018433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,balanced,0.11326400438944499
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,balanced,0.13377599914868674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,balanced,0.1548799971739451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,balanced,0.20494933923085532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,balanced,0.24865599473317465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,balanced,0.34967998663584393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,balanced,0.4322613477706909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,balanced,0.6224160194396973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,balanced,0.7888320287068685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,power_law_1.01,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,power_law_1.01,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,power_law_1.01,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,power_law_1.01,0.10611840486526489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,power_law_1.01,0.14616960287094116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,power_law_1.01,0.18789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,power_law_1.01,0.2538687944412231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,power_law_1.01,0.30690560340881345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,power_law_1.01,0.4934400081634521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,power_law_1.01,0.6242688179016114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,balanced,0.038293334345022835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,balanced,0.03386666625738144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,balanced,0.03395200024048487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,balanced,0.040949332217375435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,balanced,0.045109331607818604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,balanced,0.045007998744646706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,balanced,0.05602666735649109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,balanced,0.05657599866390228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,balanced,0.0997279981772105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,balanced,0.11383466919263203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,balanced,0.1516746679941813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,balanced,0.1819146672884623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,power_law_1.01,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,power_law_1.01,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,power_law_1.01,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,power_law_1.01,0.060121601819992064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,power_law_1.01,0.06106879711151123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,power_law_1.01,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,power_law_1.01,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,power_law_1.01,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,power_law_1.01,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,power_law_1.01,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,power_law_1.01,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,power_law_1.01,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,power_law_1.01,0.14209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,power_law_1.01,0.19259519577026368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,power_law_1.01,0.22917120456695556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,power_law_1.01,0.31943678855895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,power_law_1.01,0.39711360931396483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,power_law_1.01,0.5713215827941894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,power_law_1.01,0.7710271835327148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.0840063989162445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.07057279944419861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.1091647982597351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.110534405708313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.10277760028839111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.12033280134201049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.1595072031021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.1959488034248352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.19979519844055177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.16686079502105713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.2814527988433838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.30691840648651125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,power_law_1.2,0.41245441436767577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.20683519840240477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,power_law_1.2,0.4438784122467041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.2862400054931641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,power_law_1.2,0.7337664127349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.3984639883041382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,balanced,0.06447466711203258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,power_law_1.2,1.009113597869873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,balanced,0.0892693301041921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,balanced,0.10674666364987691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,balanced,0.10753599802652995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,balanced,0.10641599694887798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,balanced,0.10838933785756429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,balanced,0.10736533006032307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,balanced,0.10875733693440755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,balanced,0.10969600081443787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,balanced,0.1113866666952769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,balanced,0.11156266927719116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,balanced,0.11327466368675232
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,balanced,0.11718400319417317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,balanced,0.1204746663570404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,balanced,0.12395733594894409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,balanced,0.14461867014567056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,balanced,0.14511467019716898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,balanced,0.17805866400400797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,balanced,0.1880106727282206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,balanced,0.2579306761423747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,balanced,0.2407840092976888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,balanced,0.3200800021489461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,balanced,0.357589324315389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,balanced,0.49489065011342365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,balanced,0.5826026598612467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,power_law_1.2,1.4735487937927245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,power_law_1.2,1.6159103393554688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,balanced,0.02938133229811986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,balanced,0.03658133248488108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,balanced,0.04218133290608724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,balanced,0.060458665092786155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,balanced,0.07246399919191997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,balanced,0.0888213316599528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,balanced,0.12525866429011026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,balanced,0.161189337571462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,balanced,0.22538665930430093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,balanced,0.27510933081309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.10527360439300537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.11278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.1050495982170105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.09662719964981079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.10869120359420777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.11033600568771362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.10786559581756591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.11051520109176635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.11607680320739747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.11345280408859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.12014080286026001
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.12622079849243165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.13253120183944703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.151801598072052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.16387840509414672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.1935871958732605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.21880319118499755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.2924544095993042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.36211841106414794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,power_law_1.01,0.438643217086792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,power_law_1.01,0.436352014541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,power_law_1.01,0.6418047904968261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,power_law_1.01,0.8769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.04129279851913452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.0900160014629364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.10399359464645386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.13879679441452025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.17715840339660643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.24236159324645995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.3382080078125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.3601984024047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.5573184013366699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.7253632068634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,power_law_1.01,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,power_law_1.01,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,power_law_1.01,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,power_law_1.01,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,power_law_1.01,0.12768640518188476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,power_law_1.01,0.16104960441589355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,power_law_1.01,0.16741759777069093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,power_law_1.01,0.1726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,power_law_1.01,0.18208639621734618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,power_law_1.01,0.1902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,power_law_1.01,0.19593600034713746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,power_law_1.01,0.201580810546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,power_law_1.01,0.21220479011535645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,power_law_1.01,0.2194751977920532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,power_law_1.01,0.23369600772857665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,power_law_1.01,0.26732161045074465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,power_law_1.01,0.2922368049621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,power_law_1.01,0.35708160400390626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,power_law_1.01,0.40976638793945314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,power_law_1.01,0.5345024108886719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,power_law_1.01,0.6358784198760986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,power_law_1.01,0.8647616386413575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,power_law_1.01,1.0874624252319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,power_law_1.01,1.5275456428527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,power_law_1.01,1.9644351959228517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.01,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.01,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.01,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.01,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.01,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.01,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.01,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.01,0.07703040242195129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.01,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.01,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.01,0.09305599927902222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.01,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.01,0.12389119863510131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.01,0.1550655961036682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.01,0.20161919593811034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.01,0.21375999450683594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.01,0.2687551975250244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.01,0.32730879783630373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.01,0.4292736053466797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.01,0.5476352214813233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,power_law_1.2,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,power_law_1.2,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,power_law_1.2,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,power_law_1.2,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,power_law_1.2,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,power_law_1.2,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,power_law_1.2,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,power_law_1.2,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,power_law_1.2,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,power_law_1.2,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,power_law_1.2,0.09359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,power_law_1.2,0.12108160257339477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,power_law_1.2,0.1265920042991638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,power_law_1.2,0.15364480018615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,power_law_1.2,0.18587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,power_law_1.2,0.24364159107208253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,power_law_1.2,0.3116863965988159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,power_law_1.2,0.4156223773956299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,power_law_1.2,0.5563712120056152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,power_law_1.2,0.8128767967224121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,power_law_1.2,1.050175952911377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,balanced,0.12575999895731607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,balanced,0.21014932791392008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,balanced,0.3834880193074544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,balanced,0.38442134857177734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,balanced,0.3819146553675334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,balanced,0.3877706527709961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,balanced,0.39077333609263104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,balanced,0.39019731680552167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,balanced,0.3911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,balanced,0.3910133441289266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,balanced,0.39874664942423504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,balanced,0.4060426553090413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,balanced,0.40933334827423096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,balanced,0.42315200964609784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,balanced,0.45444798469543457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,balanced,0.4593386650085449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,balanced,0.4995306730270386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,balanced,0.544490655263265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,balanced,0.5876373449961344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,balanced,0.7315466403961182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,balanced,0.8249920209248861
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,balanced,1.1182400385538738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,balanced,1.4549919764200847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,balanced,2.1232426961263022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,balanced,2.8284425735473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,balanced,4.170255978902181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,balanced,5.518351872762044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,power_law_1.01,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,power_law_1.01,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,power_law_1.01,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,power_law_1.01,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,power_law_1.01,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,power_law_1.01,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,power_law_1.01,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,power_law_1.01,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,power_law_1.01,0.07315199971199035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,power_law_1.01,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,power_law_1.01,0.1280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,power_law_1.01,0.15957759618759154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,power_law_1.01,0.2139456033706665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,power_law_1.01,0.2991679906845093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,power_law_1.01,0.41146240234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,power_law_1.01,0.5194496154785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,balanced,0.027744000156720478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,balanced,0.029445332785447437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,balanced,0.03860266755024592
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,balanced,0.04152533411979675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,balanced,0.08954667051633199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,balanced,0.10845866799354553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,balanced,0.13993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,balanced,0.16335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,balanced,0.02938666691382726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,balanced,0.03260799994071325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,balanced,0.04842666784922282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,balanced,0.06960533559322357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,balanced,0.0765226682027181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,balanced,0.07897066573301952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,balanced,0.07914133369922638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,balanced,0.07810666660467784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,balanced,0.07730133334795634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,balanced,0.07916266719500224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,balanced,0.08387200037638347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,balanced,0.08558932940165202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,balanced,0.08597866694132487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,balanced,0.10642133156458537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,balanced,0.12521599729855856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,balanced,0.10784000158309937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,balanced,0.1390666663646698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,balanced,0.14686933159828186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,balanced,0.19205333789189658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,balanced,0.20762133598327637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,balanced,0.28700800736745197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,balanced,0.3298826615015666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,balanced,0.0565280020236969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,balanced,0.05783999959627787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,balanced,0.07501866420110066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,balanced,0.09889066219329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,balanced,0.10154133041699727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,balanced,0.10500799616177876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,balanced,0.10539199908574422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,balanced,0.10424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,balanced,0.10725866754849751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,balanced,0.1092746655146281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,balanced,0.11286933223406474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,balanced,0.11342400312423706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,balanced,0.12407466769218445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,balanced,0.12389866511027019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,balanced,0.1272053321202596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,balanced,0.14204266667366028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.15034133195877075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.17576533555984497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.19830399751663208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.24732265869776407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.29662932952245075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.4262133439381917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.5319946606953939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.7684640089670817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,balanced,1.0134613513946533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,balanced,0.06691733499368031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,balanced,0.08105066418647766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,balanced,0.09028266867001851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,balanced,0.09103999535242717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,balanced,0.09295466542243958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,balanced,0.09432533383369446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,balanced,0.09586133559544881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,balanced,0.10593600074450175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,balanced,0.10746133327484131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,balanced,0.12025066216786702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,balanced,0.12780800461769104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,balanced,0.13055466612180075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,balanced,0.15195199847221375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,balanced,0.16594666242599487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,balanced,0.20545599857966104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,balanced,0.23760000864664713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,balanced,0.3121066689491272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,balanced,0.39932799339294434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,balanced,0.5567040046056112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,balanced,0.7375199794769287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.09764479994773864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.14379520416259767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.24320640563964843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.34140160083770754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.44498558044433595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6114687919616699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.6448256015777588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.679795217514038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6713344097137451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7017920017242432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.731935977935791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7461887836456299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7675903797149658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.7762495994567871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.8018495559692382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.8537535667419434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,power_law_1.01,0.8928319931030273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,power_law_1.01,0.9591487884521485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.097696018218994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.1891648292541503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.412063980102539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.5263296127319337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,power_law_1.01,1.8894336700439454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.2880832672119142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.095840072631836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,power_law_1.01,3.8759296417236326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,balanced,0.0787360022465388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,balanced,0.0809333324432373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,balanced,0.0784853349129359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,balanced,0.08115200201670329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,balanced,0.08201066652933757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,balanced,0.08225066463152568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,balanced,0.08449066678682964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,balanced,0.08456533153851827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,balanced,0.08673066894213359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,balanced,0.08582400282224019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,balanced,0.08962133526802063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,balanced,0.09092266360918681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,balanced,0.09330667058626811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,balanced,0.09780800342559814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,balanced,0.10121599833170573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,balanced,0.10779199997584026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,balanced,0.11763733625411987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,balanced,0.1360160013039907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,balanced,0.15653866529464722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,balanced,0.19290133317311606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,balanced,0.2281386653582255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,balanced,0.3059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,balanced,0.3765653371810913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.11226880550384521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.11550079584121704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.11997439861297607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.12352639436721802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.1277184009552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.13080960512161255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.13466880321502686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.1423359990119934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.14636800289154053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.1566848039627075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.1861248016357422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.20434560775756835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.25445759296417236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.3004224061965942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.40872960090637206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5088640213012695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.7104576110839844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.9045503616333008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.03018240034580231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.058303999900817874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.06919040083885193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.10465919971466064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.18677120208740233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,power_law_1.01,0.2696703910827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,power_law_1.01,0.30268800258636475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.2987071990966796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,1.688742446899414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,balanced,0.050255998969078064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,balanced,0.07213333249092102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,balanced,0.09785067041714986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,balanced,0.09788266817728679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,balanced,0.09686400492986043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,balanced,0.09798399607340495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,balanced,0.09825600186983745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,balanced,0.1018453339735667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,balanced,0.10713600118954976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,balanced,0.1086186667283376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,balanced,0.11121599872907002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,balanced,0.11998933553695679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,balanced,0.1263146698474884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,balanced,0.1404853363831838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,balanced,0.15651200215021768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,balanced,0.1822506586710612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,balanced,0.21127466360727945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,balanced,0.27478400866190594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,balanced,0.32869333028793335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,balanced,0.45922664801279706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,balanced,0.564789334932963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.09534080028533935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.08017920255661011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.08019840121269226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.0873856008052826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.08830080032348633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.09571200013160705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.10049279928207397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.1534656047821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.15537919998168945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.21001598834991456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.23036799430847169
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.3271104097366333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.41521282196044923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,power_law_1.2,0.6227456092834472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,power_law_1.2,0.7442815780639649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.09240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.1344383955001831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.16993919610977173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.204256010055542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.2625344038009644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.3408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.47102718353271483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.6136960029602051
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,balanced,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,balanced,0.0505973349014918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,balanced,0.0545653353134791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,balanced,0.07073066631952922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,balanced,0.08062399923801422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,balanced,0.09920533498128255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,balanced,0.12801600495974222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,balanced,0.15747732917467752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,balanced,0.2106026609738668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,balanced,0.24849067131678262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,balanced,0.34277331829071045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,balanced,0.44198401769002277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,power_law_1.2,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,power_law_1.2,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,power_law_1.2,0.1186303973197937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,power_law_1.2,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,power_law_1.2,0.12333439588546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,power_law_1.2,0.11758719682693482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,power_law_1.2,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,power_law_1.2,0.11643520593643189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,power_law_1.2,0.1100160002708435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,power_law_1.2,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,power_law_1.2,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,power_law_1.2,0.11365760564804077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,power_law_1.2,0.1260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,power_law_1.2,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,power_law_1.2,0.141702401638031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,power_law_1.2,0.14558720588684082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,power_law_1.2,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,power_law_1.2,0.18288639783859253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,power_law_1.2,0.22076160907745362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,power_law_1.2,0.23666560649871826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,power_law_1.2,0.29656960964202883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,power_law_1.2,0.3770495891571045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,power_law_1.2,0.5012928009033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,power_law_1.2,0.6450431823730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,power_law_1.2,0.8820159912109375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,power_law_1.2,1.1093055725097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,balanced,0.04766400158405304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,balanced,0.051632001996040344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,balanced,0.05776533484458923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,balanced,0.06604800124963124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,balanced,0.0724426656961441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,balanced,0.0846613347530365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,balanced,0.09682133793830872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,balanced,0.11275733510653178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,balanced,0.1379200021425883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,balanced,0.1836479902267456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,balanced,0.22381865978240967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,balanced,0.31407999992370605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,balanced,0.38841064771016437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,balanced,0.07716266810894012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,balanced,0.10221866766611735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,balanced,0.15260266264279684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,balanced,0.22166399161020914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,balanced,0.2509546677271525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,balanced,0.25269333521525067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,balanced,0.2523840069770813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,balanced,0.2533866763114929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,balanced,0.25463465849558514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,balanced,0.253440002600352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,balanced,0.2579519947369893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,balanced,0.25918932755788165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,balanced,0.25919467210769653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,balanced,0.2662293314933777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,balanced,0.26799466212590534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,balanced,0.2720106641451518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,balanced,0.2976906696955363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,balanced,0.29208532969156903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,balanced,0.3712960084279378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,balanced,0.3435200055440267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,balanced,0.524778683980306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,balanced,0.41173334916432697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,balanced,0.6147786776224772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,balanced,0.5859626531600952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,balanced,0.9031466643015543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,balanced,0.8942506313323975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,balanced,0.13117333253224692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,balanced,0.15774400035540262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,balanced,0.15797866384188333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,balanced,0.15812800327936807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,balanced,0.15923200050989786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,balanced,0.16082666317621866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,balanced,0.16077333688735962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,balanced,0.16612799962361655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,balanced,0.16550933321317038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,balanced,0.16706132888793945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,balanced,0.17536532878875732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,balanced,0.17485332489013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,balanced,0.18315200010935465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,balanced,0.21174399058024088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,balanced,0.20744532346725464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,balanced,0.2936319907506307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,balanced,0.26868265867233276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,balanced,0.377679983774821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,balanced,0.36476266384124756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,balanced,0.5175093412399292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,balanced,0.583951989809672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,balanced,0.8331092993418375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,balanced,0.974671999613444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.0913856029510498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.09893760085105896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.10867199897766114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.11350400447845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.11277439594268798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.11420799493789673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.11567360162734985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.1161728024482727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.12970880270004273
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,power_law_1.2,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.13676799535751344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.15189759731292723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.16926079988479614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.2278656005859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.2986112117767334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.3606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.5035327911376953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.608460807800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.8689855575561524
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.1053183555603028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,power_law_1.2,0.07925119996070862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,power_law_1.2,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,power_law_1.2,0.08218880295753479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,power_law_1.2,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,power_law_1.2,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.09233919978141784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.10843520164489746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,power_law_1.2,0.09115520119667053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.13007359504699706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,power_law_1.2,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.18305920362472533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,power_law_1.2,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.20167040824890137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,power_law_1.2,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.3110464096069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,power_law_1.2,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.3180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,power_law_1.2,0.15241600275039674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.3332607984542847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,power_law_1.2,0.18099839687347413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.350598406791687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,power_law_1.2,0.23415040969848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.3584896087646484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,power_law_1.2,0.2838207960128784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.3595776081085205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.3768640041351318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,power_law_1.2,0.40793600082397463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.3771775960922241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,power_law_1.2,0.4913472175598145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.38302719593048096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.397760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,power_law_1.2,0.7351808071136474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.3975872039794922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,power_law_1.2,0.9066816329956054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.42914562225341796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.4670015811920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.5075200080871582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.6182015895843506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.6436543941497803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,0.8545087814331055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,0.9905535697937011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,1.3511167526245118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,1.6627967834472657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,2.3257984161376952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,2.7714752197265624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,balanced,0.05027199784914652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,balanced,0.057461331288019814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,balanced,0.07259200016657512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,balanced,0.07235733171304067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,balanced,0.0745066652695338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,balanced,0.08263466755549113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,balanced,0.08803199728329976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,balanced,0.09057066837946574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,balanced,0.09529067079226176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,balanced,0.09999466935793559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,balanced,0.10727999607721965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,balanced,0.12133866548538208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,balanced,0.1301759978135427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,balanced,0.1646453340848287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,balanced,0.18505599101384482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,balanced,0.2346400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,balanced,0.28148800134658813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,balanced,0.378762682278951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,balanced,0.47036266326904297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.12780799865722656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.14793599843978883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.18143359422683716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.23783679008483888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.3136127948760986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.3992831945419312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,power_law_1.01,0.535916805267334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,balanced,0.08307733138402303
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,balanced,0.10524266958236694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,balanced,0.13826666275660196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,balanced,0.2016800045967102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,balanced,0.32580800851186115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,balanced,0.5807199875513712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,balanced,0.575984001159668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,balanced,0.5779199997584025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,balanced,0.5867946545282999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,balanced,0.5855199893315634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,balanced,0.5921226739883423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,balanced,0.5995413462320963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,balanced,0.6035093466440836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,balanced,0.6139466762542725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,balanced,0.6372106472651163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,balanced,0.6419839859008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,balanced,0.6596853335698446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,balanced,0.7093919912974039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.7583200136820475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.8579253355662028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.9079519907633463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,balanced,1.068666696548462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.2386079629262288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,balanced,1.660960038503011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,balanced,1.9958346684773762
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,balanced,2.896944046020508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,balanced,3.588421185811361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,power_law_1.2,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,power_law_1.2,0.11121920347213746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,power_law_1.2,0.12398719787597656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,power_law_1.2,0.15928319692611695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,power_law_1.2,0.17559679746627807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,power_law_1.2,0.1873471975326538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,power_law_1.2,0.24410240650177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,power_law_1.2,0.2473151922225952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,power_law_1.2,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,power_law_1.2,0.2541887998580933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,power_law_1.2,0.2522239923477173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,power_law_1.2,0.25668480396270754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,power_law_1.2,0.2698879957199097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,power_law_1.2,0.28173439502716063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,power_law_1.2,0.29443199634552003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,power_law_1.2,0.304038405418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,power_law_1.2,0.3278656005859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,power_law_1.2,0.34581120014190675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,power_law_1.2,0.35304319858551025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,power_law_1.2,0.4026048183441162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,power_law_1.2,0.432096004486084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,power_law_1.2,0.5091775894165039
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,power_law_1.2,0.5516352176666259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,power_law_1.2,0.6783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,power_law_1.2,0.7709568023681641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,power_law_1.2,1.0731519699096679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,power_law_1.2,1.2782400131225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.13648639917373656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.15476479530334472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.21941120624542237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.2797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,balanced,0.06099733213583628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,balanced,0.07394133508205414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,balanced,0.07619733115037282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,balanced,0.07464533547560374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,balanced,0.07788266738255818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,balanced,0.07549333572387695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,balanced,0.07698133091131847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,balanced,0.07643733421961467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,balanced,0.07684266567230225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,balanced,0.07805333534876506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,balanced,0.07861333092053731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,balanced,0.08654399712880452
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08709866801897685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,balanced,0.09213333328564961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.13201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.15214932958285013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.1962826649347941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.2371573249499003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.3189599911371867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.40910398960113525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.5794666608174642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,balanced,0.7440053621927897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,power_law_1.2,0.4024384021759033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,power_law_1.2,0.4842239856719971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.2326591968536377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.27861120700836184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.371724796295166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.42856321334838865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.5185728073120117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.6499904155731201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.7225535869598388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.6753151893615723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.7544832229614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.6945280075073242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.6970816135406495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.7400832176208496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,power_law_1.2,0.7601535797119141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,power_law_1.2,0.7729599952697754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,power_law_1.2,0.7998335838317872
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,power_law_1.2,0.8743231773376465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,power_law_1.2,0.9496383666992188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,power_law_1.2,0.9423871994018554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.1065279960632324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,power_law_1.2,1.0953472137451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,power_law_1.2,1.211673641204834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,power_law_1.2,1.3424960136413575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,power_law_1.2,1.7024063110351562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,power_law_1.2,1.9239999771118164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,power_law_1.2,2.6105152130126954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,power_law_1.2,3.1169855117797853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,balanced,0.04206933577855428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,balanced,0.07593599955240886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,balanced,0.11441066861152649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,balanced,0.11389333009719849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,balanced,0.11354133486747742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,balanced,0.11167466640472412
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,balanced,0.11313600341478984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,balanced,0.1122826635837555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,balanced,0.11518399914105733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,balanced,0.11405332883199056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,balanced,0.11507200201352437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,balanced,0.12061333656311035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,balanced,0.12115200360616048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,balanced,0.12238400181134541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,balanced,0.12779733538627625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,balanced,0.1306826670964559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,balanced,0.1399999956289927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,balanced,0.14940266807874045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,balanced,0.17499732971191406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,balanced,0.18582934141159058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,balanced,0.23274133602778116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,balanced,0.2663840055465698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,balanced,0.34889066219329834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,balanced,0.4284106492996216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,power_law_1.2,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,power_law_1.2,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,power_law_1.2,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,power_law_1.2,0.09051520228385926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,power_law_1.2,0.09299839735031128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,power_law_1.2,0.09247999787330627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,power_law_1.2,0.09340800046920776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,power_law_1.2,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,power_law_1.2,0.09683840274810791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,power_law_1.2,0.10232959985733033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,power_law_1.2,0.10845439434051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,power_law_1.2,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,power_law_1.2,0.13078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.2,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.2,0.23269760608673096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.2,0.29665279388427734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.2,0.38480639457702637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.2,0.5165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.2,0.6540927886962891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.10579839944839478
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.1451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.2,1.1341631889343262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.16016000509262085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.22548480033874513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.2828927993774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.2,1.4507455825805664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.399782395362854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.5153855800628662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.08636800050735474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.09507200121879578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.11788159608840942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.13640960454940795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.2050112009048462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.21337599754333497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.21712639331817626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.22248320579528807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.23092479705810548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.23758718967437745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.24401919841766356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.25569920539855956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.2640000104904175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.2708224058151245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.2927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.3403903961181641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.3812671899795532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.47036161422729494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.5533504009246826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.7218944072723389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,0.8878463745117188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.2255104064941407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,1.5699328422546386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.2,2.245657539367676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.2,2.917631912231445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.13052159547805786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.1743551969528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.2194432020187378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.2802175998687744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.4869056224822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.4973504066467285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.5212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.5099584102630615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.5374847888946533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.5494016170501709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.5688064098358154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.5540863990783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.5751616001129151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.5951104164123535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.6049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.6395199775695801
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.7017151832580566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,power_law_1.2,0.7412799835205078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,power_law_1.2,0.8492544174194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,power_law_1.2,0.9813504219055176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,power_law_1.2,1.1660096168518066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,power_law_1.2,1.1360511779785156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,power_law_1.2,1.4869440078735352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,power_law_1.2,1.6527807235717773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.0864575982093811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.10869120359420777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.11791360378265381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.10808320045471191
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,power_law_1.2,2.253343963623047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.12503039836883545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.1263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.13108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.13099520206451415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.14792959690093993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.16283520460128784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.1731775999069214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.20662400722503663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.22650880813598634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.2804352045059204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.35770881175994873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.4624000072479248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,power_law_1.2,0.5807231903076172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,power_law_1.2,0.8071552276611328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,power_law_1.2,1.0601152420043944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,power_law_1.2,2.961689567565918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,power_law_1.01,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,power_law_1.01,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,power_law_1.01,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,power_law_1.01,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,power_law_1.01,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,power_law_1.01,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,power_law_1.01,0.04815999865531921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,power_law_1.01,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,power_law_1.01,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,power_law_1.01,0.15916800498962402
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,power_law_1.01,0.19941760301589967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,power_law_1.01,0.288428807258606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,power_law_1.01,0.36243839263916017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.07790719866752624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.09221760034561158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.09691519737243652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.10186879634857178
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.10592000484466553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.1086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.12645119428634644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.13049600124359131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.14502400159835815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.16324479579925538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.1921023964881897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.2356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.3038655996322632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.3742847919464111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.5032256126403809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.6071360111236572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.8827712059020996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.294713592529297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.08659840226173401
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.10844160318374634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.16640640497207643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.20864639282226563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.26518399715423585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.373471999168396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.4810624122619629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,balanced,0.051674668987592064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,balanced,0.07779199878374736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,balanced,0.11072533329327901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,balanced,0.14222400387128195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,balanced,0.1418880025545756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,balanced,0.1425920029481252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,balanced,0.1437440017859141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,balanced,0.14528000354766846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,balanced,0.14999999602635702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,balanced,0.15041066209475198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,balanced,0.15264000495274863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,balanced,0.16033599774042764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,balanced,0.16262400150299072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,balanced,0.16687466700871786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,balanced,0.18380266427993774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,balanced,0.19960000117619833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,balanced,0.2218666672706604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,balanced,0.24905065695444742
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,balanced,0.2998453378677368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,balanced,0.36931200822194415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,balanced,0.4715253512064616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,balanced,0.596399982770284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,balanced,0.811562697092692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,balanced,1.0463573137919109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,power_law_1.2,0.5710847854614258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,balanced,0.029477333029111225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,balanced,0.05876799921194712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,balanced,0.06118399898211161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,balanced,0.06113066772619883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,balanced,0.06717333197593689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,balanced,0.06699733436107635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,balanced,0.06809066732724507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,balanced,0.07329600056012471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,balanced,0.08272000153859456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,balanced,0.10660266876220703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,balanced,0.11658133069674174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,balanced,0.15611732999483743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,balanced,0.1771679917971293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,balanced,0.24723732471466064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,balanced,0.27694932619730633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,power_law_1.2,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,power_law_1.2,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,power_law_1.2,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,power_law_1.2,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,power_law_1.2,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,power_law_1.2,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,power_law_1.2,0.14302719831466676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,power_law_1.2,0.20420479774475098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,power_law_1.2,0.27198081016540526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,power_law_1.2,0.39628798961639405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,balanced,0.05426133175690969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,balanced,0.054383998115857445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,balanced,0.05547733108202616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,balanced,0.05644799768924713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,balanced,0.05840000013510386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,balanced,0.06506133576234181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,balanced,0.06844800213972728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,balanced,0.08454933762550354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,balanced,0.08404800295829773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,balanced,0.11182933052380879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,balanced,0.12557333707809448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,balanced,0.18306134144465128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,balanced,0.18084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,balanced,0.24040534098943075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,balanced,0.2694293260574341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,balanced,0.3657013177871704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,balanced,0.4463786681493123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,power_law_1.2,0.47399678230285647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,balanced,0.0883733332157135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,balanced,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,balanced,0.09341333309809367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,balanced,0.09332266449928284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,balanced,0.09558399518330891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,balanced,0.09753066301345825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,balanced,0.09822400410970052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,balanced,0.10441600282986958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,balanced,0.1148426632086436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,balanced,0.11789333820343018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,balanced,0.13359466195106506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,balanced,0.13990400234858194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,balanced,0.17874133586883545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,balanced,0.2039573391278585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,balanced,0.2745973269144694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,balanced,0.3131893277168274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,balanced,0.45491735140482586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,balanced,0.5571253299713135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,balanced,0.7762986818949381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,balanced,1.0134773254394531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,balanced,0.058330665032068886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,balanced,0.08787733316421509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,balanced,0.12078400452931722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,balanced,0.18036800622940063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,balanced,0.29923200607299805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,balanced,0.41765864690144855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,balanced,0.4187946716944377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,balanced,0.41963199774424237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,balanced,0.4208746751149495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,balanced,0.42420268058776855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,balanced,0.42772801717122394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,balanced,0.43165866533915204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,balanced,0.43164801597595215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,balanced,0.4427359898885091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,balanced,0.44442133108774823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,balanced,0.4572533369064331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,balanced,0.46674132347106934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.48148266474405926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.5128480195999146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.5394506851832072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.6131786505381266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.6794133186340332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.8040693600972494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,balanced,0.9815733432769775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,balanced,1.2202133337656658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,balanced,1.6393760045369465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,power_law_1.01,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,power_law_1.01,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,power_law_1.01,0.09118080139160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,power_law_1.01,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,power_law_1.01,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,power_law_1.01,0.09763200283050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,power_law_1.01,0.10149760246276855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,power_law_1.01,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,power_law_1.01,0.10476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,power_law_1.01,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,power_law_1.01,0.11853439807891845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,power_law_1.01,0.12049280405044556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,power_law_1.01,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.01,0.15292160511016845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.01,0.18513920307159423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.01,0.21580159664154053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.01,0.2830399990081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.01,0.3405247926712036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.01,0.48090238571166993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.01,0.6311039924621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.01,0.8414143562316895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.01,1.1341312408447266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.18620799779891967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.25121920108795165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.3268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.3772671937942505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.4490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.4662144184112549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.4655935764312744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.5176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.49584641456604006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.5070911884307862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.5578815937042236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.5655039787292481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.5830080032348632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.5997312068939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.6494783878326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.7108096122741699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,power_law_1.2,0.7869440078735351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,power_law_1.2,0.867251205444336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.0285183906555175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.1440704345703125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,power_law_1.2,1.4207039833068849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,power_law_1.2,1.5993791580200196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,power_law_1.2,2.153651237487793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,power_law_1.2,2.6625471115112305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,balanced,0.03878933439652125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,balanced,0.046821330984433494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.06863999863465627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.10839466253916423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.1241386632124583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.1611840029557546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.19381866852442423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.27718400955200195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.3564693530400594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,balanced,0.032469332218170166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,balanced,0.03198933353026708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,balanced,0.04454400142033895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,balanced,0.045466666420300804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,balanced,0.06108266611893972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,balanced,0.06712000072002411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,balanced,0.09497599800427754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,balanced,0.12378666798273723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,balanced,0.1479573349157969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,balanced,0.19947733481725058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,balanced,0.2444159984588623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.2,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.2,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.2,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.2,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.2,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.2,0.04821760058403015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.2,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.2,0.1280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.2,0.17589759826660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.2,0.2842751979827881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,balanced,0.0360000009338061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,balanced,0.04855466882387797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,balanced,0.04948266843954722
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,balanced,0.049600000182787575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.2,0.26448640823364256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,balanced,0.06421866516272227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,balanced,0.07021333277225494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,balanced,0.08413867155710857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,balanced,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,balanced,0.12827733159065247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,balanced,0.1481013298034668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,balanced,0.20147732893625894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,balanced,0.25886933008829754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,balanced,0.36580801010131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,balanced,0.4665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,balanced,0.03178666780392329
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,balanced,0.03271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,balanced,0.03312533348798752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,balanced,0.0349440003434817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,balanced,0.0421013335386912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,balanced,0.045408000548680626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,balanced,0.05782933533191681
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,balanced,0.06994133194287618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,balanced,0.09181867043177287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,balanced,0.10033599535624187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,balanced,0.13645866513252258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,balanced,0.16862932840983072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,balanced,0.0296426663796107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,balanced,0.03422400106986364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,balanced,0.03603200117746989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,balanced,0.05458133419354757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,balanced,0.10749866565068562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,balanced,0.13613333304723105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,balanced,0.17917333046595255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,balanced,0.21953600645065308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,power_law_1.2,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,power_law_1.2,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,power_law_1.2,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,power_law_1.2,0.05281280279159546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,power_law_1.2,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,power_law_1.2,0.11002880334854126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,power_law_1.2,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,power_law_1.2,0.15354880094528198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,power_law_1.2,0.2167680025100708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,power_law_1.2,0.26627840995788576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,power_law_1.2,0.41904001235961913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,power_law_1.2,0.4767104148864746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.06090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.08113279938697815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.10249600410461426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.13370239734649658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.16469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.20347518920898439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.28219521045684814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.3441728115081787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.45759358406066897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.633625602722168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.2,0.04398080110549927
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.2,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.2,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.2,0.07240960001945496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.2,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.2,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.2,0.09208319783210754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.2,0.09719679951667785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.2,0.10845439434051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.2,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.2,0.13493119478225707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.2,0.14010239839553834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.2,0.14736000299453736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.2,0.18182400465011597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.2,0.2340480089187622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.2,0.2837631940841675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.2,0.4826367855072021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.2,0.4564671993255615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,power_law_1.01,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,power_law_1.01,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,power_law_1.01,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,power_law_1.01,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,power_law_1.01,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,power_law_1.01,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,power_law_1.01,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,power_law_1.01,0.11337599754333497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,power_law_1.01,0.13955199718475342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,power_law_1.01,0.19166719913482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,power_law_1.01,0.2659712076187134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,power_law_1.01,0.3640255928039551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,power_law_1.01,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,power_law_1.01,0.13180160522460938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,power_law_1.01,0.11662720441818238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,power_law_1.01,0.14802559614181518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,power_law_1.01,0.13800959587097167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,power_law_1.01,0.16281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,power_law_1.01,0.16505600214004518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,power_law_1.01,0.17161600589752196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,power_law_1.01,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,power_law_1.01,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,power_law_1.01,0.17868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,power_law_1.01,0.17466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,power_law_1.01,0.18019200563430787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,power_law_1.01,0.19527679681777954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,power_law_1.01,0.2033344030380249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,power_law_1.01,0.20784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,power_law_1.01,0.2217855930328369
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,power_law_1.01,0.24469759464263915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,power_law_1.01,0.26728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,power_law_1.01,0.31719040870666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,power_law_1.01,0.37731199264526366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,power_law_1.01,0.47574400901794434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,power_law_1.01,0.5177152156829834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,power_law_1.01,0.6098944187164307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,power_law_1.01,0.8031423568725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,power_law_1.01,1.0694656372070312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,power_law_1.01,1.332313632965088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,power_law_1.01,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,power_law_1.01,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,power_law_1.01,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,power_law_1.01,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,power_law_1.01,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,power_law_1.01,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,power_law_1.01,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,power_law_1.01,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,power_law_1.01,0.06782079935073852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,power_law_1.01,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,power_law_1.01,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,power_law_1.01,0.13381119966506957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,power_law_1.01,0.15902719497680665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,power_law_1.01,0.2297215938568115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,power_law_1.01,0.2893824100494385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,balanced,0.08039466540018718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,balanced,0.0981333355108897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,balanced,0.11282133062680562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,balanced,0.11363200346628825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,balanced,0.1122826635837555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,balanced,0.11469333370526631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,balanced,0.11514133214950562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,balanced,0.11715199549992879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,balanced,0.11762133240699768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,balanced,0.11936533451080322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,balanced,0.12704533338546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,balanced,0.13051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,balanced,0.13569066921869913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,balanced,0.1428053379058838
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,balanced,0.15154133240381876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,balanced,0.1732906699180603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,balanced,0.1918026606241862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,balanced,0.24050132433573404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,balanced,0.28947200377782184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,balanced,0.3920266628265381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,balanced,0.5252853234608968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,balanced,0.7240266799926758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,balanced,0.9437013467152914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,power_law_1.01,0.41047039031982424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,power_law_1.01,0.4544064044952393
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,power_law_1.01,0.7306047916412354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,power_law_1.01,1.0187840461730957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.06844159960746765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.0861952006816864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.13333760499954223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.14816639423370362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.1898751974105835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.23447039127349853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.29541759490966796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.36947200298309324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.5332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.0764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.08764799833297729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.09726079702377319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.12565120458602905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.14866559505462645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.15406719446182252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.1907263994216919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.21992321014404298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,power_law_1.2,0.29564800262451174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,power_law_1.2,0.4180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,power_law_1.2,0.6293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.68787841796875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,power_law_1.2,0.7509759902954102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.06046079993247986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.0866752028465271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.1028480052947998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.1215999960899353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.1931391954421997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.23176319599151612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.31751039028167727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.40953598022460935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,power_law_1.01,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,power_law_1.01,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,power_law_1.01,0.05729280114173889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,power_law_1.01,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,power_law_1.01,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,power_law_1.01,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,power_law_1.01,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,power_law_1.01,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,power_law_1.01,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,power_law_1.01,0.10161279439926148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,power_law_1.01,0.1026304006576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,power_law_1.01,0.11564799547195434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,power_law_1.01,0.12333439588546753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,power_law_1.01,0.1374848008155823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,power_law_1.01,0.16387840509414672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,power_law_1.01,0.16779520511627197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,power_law_1.01,0.19962879419326782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,power_law_1.01,0.24625918865203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,power_law_1.01,0.32083840370178224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,power_law_1.01,0.40163841247558596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,power_law_1.01,0.5521152019500732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,power_law_1.01,0.7127359867095947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,power_law_1.01,1.0139904022216797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,power_law_1.01,1.3186944007873536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,power_law_1.01,0.060627198219299315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,power_law_1.01,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,power_law_1.01,0.06362879872322083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,power_law_1.01,0.06721280217170715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,power_law_1.01,0.06949759721755981
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,power_law_1.01,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,power_law_1.01,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,power_law_1.01,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,power_law_1.01,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,power_law_1.01,0.12231680154800414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,power_law_1.01,0.1540992021560669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,power_law_1.01,0.18120959997177125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,power_law_1.01,0.23367040157318114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,power_law_1.01,0.29395198822021484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,power_law_1.01,0.40842881202697756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,power_law_1.01,0.5102335929870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,power_law_1.01,0.7369279861450195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,power_law_1.01,0.9348799705505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.10620800256729127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.1422592043876648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.1740224003791809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.23840000629425048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.31899518966674806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.31845760345458984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.33953280448913575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.350598406791687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.352128005027771
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.36313600540161134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.3589440107345581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.371232008934021
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.3791743993759155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.4009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.4182784080505371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.4035520076751709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.45403518676757815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.46326398849487305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.5197504043579102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,power_law_1.01,0.5447999954223632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,power_law_1.01,0.660755205154419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,power_law_1.01,0.7723008155822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,power_law_1.01,0.9478336334228515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,power_law_1.01,1.1603008270263673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,power_law_1.01,1.465932846069336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,power_law_1.01,1.9208127975463867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,balanced,0.05909866591294607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,balanced,0.08922666311264038
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,balanced,0.13405332962671915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,balanced,0.15668267011642456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,balanced,0.1553759972254435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,balanced,0.15825066963831583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,balanced,0.1599146624406179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,balanced,0.16181866327921549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,balanced,0.1637493371963501
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,balanced,0.1681386629740397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,balanced,0.17122133572896323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,balanced,0.17497066656748453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,balanced,0.18539732694625854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,balanced,0.18819733460744223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,balanced,0.19962133963902792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,balanced,0.24016533295313516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,balanced,0.26260799169540405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,balanced,0.37096532185872394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,balanced,0.39848001797993976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,balanced,0.5925439993540446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,balanced,0.6657013495763143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,balanced,1.0078240235646565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,balanced,1.1926559607187908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,balanced,1.8292746543884277
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,balanced,2.244394620259603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,balanced,0.06693333387374878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,balanced,0.09250133236249287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,balanced,0.1466506620248159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,balanced,0.2452639937400818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,balanced,0.44468267758687335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,balanced,0.6407839854558309
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,balanced,0.6374719937642416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,balanced,0.6387466589609782
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,balanced,0.6424160003662109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,balanced,0.6412320137023926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,balanced,0.6509973208109537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,balanced,0.6520640055338541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,balanced,0.6572426557540894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,balanced,0.6703519821166992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,balanced,0.6728746891021729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,balanced,0.6899627049763998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,balanced,0.7146133581797282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,balanced,0.7345120112101237
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,balanced,0.7787840366363525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,balanced,0.8204106489817301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,balanced,0.9146133263905843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,balanced,0.9851040045420328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,balanced,1.1548906962076824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,balanced,1.412218729654948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,balanced,1.8036425908406575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,balanced,2.2762719790140786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,balanced,0.03389333436886469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,balanced,0.0295413335164388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,balanced,0.058677335580190025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,balanced,0.05669333537419637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,balanced,0.056789333621660866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,balanced,0.05851200222969055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,balanced,0.05891199906667074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,balanced,0.06279466549555461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,balanced,0.06467733283837636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,balanced,0.06482133269309998
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,balanced,0.07682666679223378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,balanced,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,balanced,0.09352533022562663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,balanced,0.12572800119717917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,balanced,0.13860799868901572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,balanced,0.19050665696461996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,balanced,0.19869865973790488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.01,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.01,0.10652159452438355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.01,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.01,0.11311999559402466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.01,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.01,0.11997439861297607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.01,0.11703679561614991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.01,0.11724159717559815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.01,0.12334719896316529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.01,0.12616959810256959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.01,0.12936320304870605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.01,0.1412287950515747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.01,0.15578240156173706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.01,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.01,0.17657599449157715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.01,0.19527679681777954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.01,0.2313472032546997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.01,0.2479167938232422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.01,0.31571838855743406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.01,0.3764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.01,0.49521918296813966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.01,0.6316544055938721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,balanced,0.05365333457787832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,balanced,0.07543466488520305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,balanced,0.08038400113582611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,balanced,0.07898666461308797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,balanced,0.08101866642634074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,balanced,0.08293333152929942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,balanced,0.084197332461675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,balanced,0.08474666873613994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,balanced,0.09684800108273824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,balanced,0.11147200067838033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.12083199620246887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.13979199528694153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.1637279987335205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.19318934281667074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.21759466330210367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.29954665899276733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.3701440095901489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.5277813275655111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.6731306711832682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,power_law_1.2,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,power_law_1.2,0.05076479911804199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,power_law_1.2,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,power_law_1.2,0.0630016028881073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,power_law_1.2,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,power_law_1.2,0.09688959717750549
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.2,0.10211199522018433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.2,0.12863999605178833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.2,0.14690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.2,0.19786880016326905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.2,0.23208959102630616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.2,0.3784192085266113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.2,0.4440000057220459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.2,0.6744639873504639
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.2,0.977292823791504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.01,0.18755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.01,0.3262847900390625
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.01,0.26289279460906984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.01,0.32731521129608154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.01,0.41472640037536623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.01,0.43970561027526855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.01,0.4055935859680176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.01,0.38937599658966066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.01,0.4125631809234619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.01,0.40142078399658204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.01,0.4094207763671875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.01,0.4053055763244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.01,0.41607041358947755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.01,0.42709760665893554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.01,0.4352896213531494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.01,0.453542423248291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.01,0.4802432060241699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.01,0.5245759963989258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.01,0.5446464061737061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.01,0.6583871841430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.01,0.7510655879974365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.01,0.8987903594970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.01,0.8965503692626953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.01,1.0591808319091798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.01,1.4196800231933593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.01,1.8861440658569335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.01,2.1162111282348635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.07117440104484558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.10196479558944702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.12944639921188356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.19790719747543334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.2131648063659668
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.28168320655822754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.37837440967559816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,power_law_1.2,0.5497983932495117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,power_law_1.2,0.7255231857299804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,balanced,0.05942399799823761
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,balanced,0.06483733157316844
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,balanced,0.09630399942398071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,balanced,0.10360532999038696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,balanced,0.10368000467618306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,balanced,0.11102400223414104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,balanced,0.11153067151705424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,balanced,0.1116480032602946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,balanced,0.11351466178894043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,balanced,0.11334400375684102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,balanced,0.11496000488599141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,balanced,0.1166986624399821
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,balanced,0.1236799955368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,balanced,0.1255519986152649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,balanced,0.13243732849756876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,balanced,0.14479466279347739
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,balanced,0.15467733144760132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,balanced,0.18041066328684488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,balanced,0.20549867550532022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,balanced,0.2701813379923503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,balanced,0.28280532360076904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,balanced,0.36635732650756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,balanced,0.4349173307418823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,balanced,0.6182986497879028
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,balanced,0.7775519688924154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,power_law_1.01,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,power_law_1.01,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,power_law_1.01,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,power_law_1.01,0.030982398986816408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,power_law_1.01,0.031455999612808226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,power_law_1.01,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,power_law_1.01,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,power_law_1.01,0.1302847981452942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,power_law_1.01,0.19729280471801758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,power_law_1.01,0.24997758865356445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.1566975951194763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.20920319557189943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.3185983896255493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.41147518157958984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.4637184143066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.5866496086120605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.6194176197052002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.6382847785949707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.6603903770446777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.6810944080352783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.7030272006988525
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.7293951988220215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.753708791732788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.7679935932159424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.8049023628234864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.8899071693420411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.9602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.0938559532165528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.1599871635437011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,1.3905920028686523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,1.6514751434326171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,2.1160640716552734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.01,2.544767951965332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.01,3.394784164428711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.01,4.180742263793945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,power_law_1.2,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,power_law_1.2,0.06600319743156433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,power_law_1.2,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,power_law_1.2,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,power_law_1.2,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,power_law_1.2,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,power_law_1.2,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,power_law_1.2,0.08015999794006348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,power_law_1.2,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,power_law_1.2,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,power_law_1.2,0.1005247950553894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,power_law_1.2,0.11745920181274414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,power_law_1.2,0.12705279588699342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,power_law_1.2,0.15637120008468627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,power_law_1.2,0.1865280032157898
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,power_law_1.2,0.2330944061279297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,power_law_1.2,0.2926975965499878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,power_law_1.2,0.4031424045562744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,power_law_1.2,0.5351871967315673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,power_law_1.2,0.7425216197967529
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,power_law_1.2,1.0529855728149413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.01,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.01,0.1265727996826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.01,0.1435520052909851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.01,0.19615360498428344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.01,0.24495360851287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.01,0.35602560043334963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.01,0.3664576053619385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.01,0.3824512004852295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.01,0.39838719367980957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.01,0.4031231880187988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.01,0.4271232128143311
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.01,0.4188416004180908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.01,0.434003210067749
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.01,0.4354879856109619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.01,0.4571392059326172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.01,0.4655168056488037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.01,0.4872896194458008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.01,0.5410624027252198
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.01,0.5844223976135254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.01,0.6913280010223388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.01,0.7625855922698974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.01,0.9533760070800781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.01,1.0526464462280274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.01,1.3675904273986816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.01,1.7735040664672852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.01,2.414681625366211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.01,2.9195840835571287
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,balanced,0.05571199953556061
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,balanced,0.058517331878344216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,balanced,0.05659733215967814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,balanced,0.06041066845258077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,balanced,0.06418666740258534
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,balanced,0.06865066786607106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,balanced,0.07750399907430013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,balanced,0.09674132863680522
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,balanced,0.1148373285929362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,balanced,0.13263466954231262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,balanced,0.16720000902811685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,balanced,0.18143999576568604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,balanced,0.27158933877944946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,balanced,0.32234134276707965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,balanced,0.46346131960550946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,balanced,0.5940159956614176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,balanced,0.043237333496411644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,balanced,0.04408533374468485
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,balanced,0.0688266654809316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,balanced,0.0900426705678304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,balanced,0.10532266894976298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,balanced,0.13556800285975137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,balanced,0.16598400473594666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,balanced,0.22220265865325928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,balanced,0.2707786758740743
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,balanced,0.04725866516431173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,balanced,0.05870933334032694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,balanced,0.059392000238100685
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,balanced,0.05834133426348368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,balanced,0.059343998630841575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,balanced,0.06070933242638906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,balanced,0.06074133515357971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,balanced,0.06211733321348826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,balanced,0.06870399912198384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,balanced,0.06855466465155284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,balanced,0.07070933282375336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,balanced,0.08293333152929942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,balanced,0.07814399898052216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,balanced,0.0976586639881134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,balanced,0.11946133772532146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,balanced,0.1378986636797587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,balanced,0.18730666240056357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,balanced,0.2065546711285909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,balanced,0.29440534114837646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,balanced,0.3533173402150472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,balanced,0.0701333334048589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,balanced,0.07051733136177063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,balanced,0.06762133538722992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,balanced,0.0678719977537791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,balanced,0.06913599868615468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,balanced,0.07020799815654755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,balanced,0.07018666466077168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,balanced,0.0780266672372818
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,balanced,0.07688533266385396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,balanced,0.09011733531951904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,balanced,0.11224533120791118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,balanced,0.12006400028864543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,balanced,0.1584160029888153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,balanced,0.1780959963798523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,balanced,0.2474986712137858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,balanced,0.28968532880147296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,balanced,0.39683731396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,balanced,0.49904000759124756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,power_law_1.2,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,power_law_1.2,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,power_law_1.2,0.10163840055465698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,power_law_1.2,0.12366080284118652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,power_law_1.2,0.14477440118789672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,power_law_1.2,0.15237120389938355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,power_law_1.2,0.18063360452651978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,power_law_1.2,0.17677439451217652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,power_law_1.2,0.18483200073242187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,power_law_1.2,0.17474559545516968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,power_law_1.2,0.20561280250549316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,power_law_1.2,0.19027199745178222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,power_law_1.2,0.19631999731063843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,power_law_1.2,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,power_law_1.2,0.22291839122772217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,power_law_1.2,0.21688320636749267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,power_law_1.2,0.2531775951385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,power_law_1.2,0.2739583969116211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,power_law_1.2,0.3318592071533203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,power_law_1.2,0.3683135986328125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,power_law_1.2,0.47297282218933107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,power_law_1.2,0.5786496162414551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,power_law_1.2,0.763865613937378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,power_law_1.2,0.9235584259033203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,power_law_1.2,1.358841609954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,power_law_1.2,1.6447168350219727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,power_law_1.2,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,power_law_1.2,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,power_law_1.2,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,power_law_1.2,0.09569919705390931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,power_law_1.2,0.10929280519485474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,power_law_1.2,0.10926079750061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,power_law_1.2,0.10702719688415527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,power_law_1.2,0.11013120412826538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,power_law_1.2,0.1120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,power_law_1.2,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,power_law_1.2,0.11783039569854736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,power_law_1.2,0.12162560224533081
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,power_law_1.2,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,power_law_1.2,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,power_law_1.2,0.13671040534973145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.2,0.15482239723205565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.2,0.18152320384979248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.2,0.21559679508209229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.2,0.2662015914916992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.2,0.3402688026428223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.2,0.4540863990783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.2,0.6192575931549072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.2,0.9255807876586915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.2,1.1876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,balanced,0.03793066740036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,balanced,0.07242133220036824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,balanced,0.11403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,balanced,0.12249066432317098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,balanced,0.12384532888730367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,balanced,0.1251413325468699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,balanced,0.12389333049456279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,balanced,0.1241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,balanced,0.12489599982897441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,balanced,0.12462932864824931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,balanced,0.1271519958972931
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,balanced,0.1274133324623108
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,balanced,0.13006933530171713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,balanced,0.13246400157610574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,balanced,0.1351040005683899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,balanced,0.15732266505559286
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,balanced,0.1393333375453949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,balanced,0.16571199893951416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,balanced,0.15934399763743082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,balanced,0.21340266863505045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,balanced,0.20204265912373862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,balanced,0.2885439991950989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,balanced,0.29259733359018963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,balanced,0.4049653212229411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,balanced,0.42710399627685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,balanced,0.051311999559402466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,balanced,0.06750933329264323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,balanced,0.07382399837176006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,balanced,0.07561066746711731
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,balanced,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,balanced,0.07584533095359802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,balanced,0.07608533402283986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,balanced,0.07894933223724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,balanced,0.0795306662718455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,balanced,0.07798933486143748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,balanced,0.08473066488901775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,balanced,0.08653866251309712
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,balanced,0.0895253320535024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.10125333070755005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.11979732910792033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.12727999687194824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.15738133589426676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.17200533548990884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.22803199291229248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.26793066660563153
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.362768014272054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.45604264736175537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,balanced,0.0484799991051356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,balanced,0.05557866891225179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,balanced,0.06822933256626129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,balanced,0.08659199873606364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,balanced,0.11154666543006897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,balanced,0.1365653375784556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,balanced,0.18661866585413614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,balanced,0.2598399917284648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,balanced,0.3763146797815959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,balanced,0.4490559895833333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.0686464011669159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.11660799980163575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.13492480516433716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.17964799404144288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.19225599765777587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.28657920360565187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.35019519329071047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.4755072116851807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.611897611618042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.12977919578552247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.13788800239562987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.1841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.21381120681762694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.19688960313796997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.21886720657348632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.2551359891891479
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.2274240016937256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.23098878860473632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.2595200061798096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.23500161170959472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.24567039012908937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,balanced,0.11386666695276897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,balanced,0.11522666613260905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,balanced,0.10990400115648906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,balanced,0.11358400185902913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,balanced,0.11168000102043152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,balanced,0.11315733194351196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,balanced,0.1130560040473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,balanced,0.1149013340473175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,balanced,0.11488533020019531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,balanced,0.12151466806729634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,balanced,0.12143466869990031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,balanced,0.12358400225639343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,balanced,0.13038933277130127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,balanced,0.13409066200256348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,balanced,0.1507253348827362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,balanced,0.15625600020090738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,balanced,0.18461867173512778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,balanced,0.19988266626993814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,balanced,0.2544586658477783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.25966079235076905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,balanced,0.30034132798512775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,balanced,0.41236265500386554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,balanced,0.4967893362045288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.3151871919631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.3654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.39432320594787595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.37676799297332764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,power_law_1.2,0.4222847938537598
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,power_law_1.2,0.4909823894500732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,power_law_1.2,0.6580480098724365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,power_law_1.2,0.7348671913146972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,power_law_1.2,1.2237631797790527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,power_law_1.2,1.315174388885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,power_law_1.2,2.25467529296875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,power_law_1.2,2.5931264877319338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,power_law_1.2,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,power_law_1.2,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,power_law_1.2,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,power_law_1.2,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,power_law_1.2,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,power_law_1.2,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,power_law_1.2,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,power_law_1.2,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,power_law_1.2,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.2,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.2,0.09763200283050537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.2,0.21560959815979003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.2,0.2391040086746216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.2,0.36769280433654783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.2,0.4430975914001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,power_law_1.2,4.103488159179688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,power_law_1.2,5.820300674438476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,power_law_1.01,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,power_law_1.01,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,power_law_1.01,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,power_law_1.01,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,power_law_1.01,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,power_law_1.01,0.04839679896831513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,power_law_1.01,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,power_law_1.01,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,power_law_1.01,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.01,0.08319360017776489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.01,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.01,0.1633344054222107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.01,0.19697279930114747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.01,0.28368639945983887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.01,0.3435391902923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,balanced,0.07309333483378093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,balanced,0.09925333658854167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,balanced,0.15455466508865356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,balanced,0.2609013319015503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,balanced,0.31487466891606647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,balanced,0.3132266600926717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,balanced,0.3200693329175313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,balanced,0.31678932905197144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,balanced,0.3229973316192627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,balanced,0.31497599681218463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,balanced,0.31756800413131714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,balanced,0.31736000378926593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,balanced,0.31858666737874347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,balanced,0.3255733251571655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,balanced,0.3240639964739482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,balanced,0.3306933244069417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,balanced,0.37465067704518634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,balanced,0.3450293143590291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,balanced,0.44086400667826336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,balanced,0.3848479986190796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,balanced,0.6222293376922607
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,balanced,0.4800586700439453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,balanced,0.7334613005320231
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,balanced,0.6886773109436035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,balanced,0.995749314626058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,balanced,1.1850132942199707
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,power_law_1.01,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,power_law_1.01,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,power_law_1.01,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,power_law_1.01,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,power_law_1.01,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,power_law_1.01,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,power_law_1.01,0.10430079698562622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,power_law_1.01,0.12081279754638671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,power_law_1.01,0.17408000230789183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.01,0.20363519191741944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.01,0.2712127923965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.01,0.3500735998153687
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,balanced,0.03994666785001755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,balanced,0.04345066845417023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,balanced,0.047557334105173744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,balanced,0.04783466458320618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,balanced,0.05027733246485392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,balanced,0.05574400226275126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,balanced,0.0911253293355306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,balanced,0.11513066291809082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,balanced,0.1432319978872935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,balanced,0.18879467248916626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,balanced,0.2313013275464376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.01,0.5642303943634033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.01,0.6854335784912109
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.01,1.0329792022705078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.01,1.1898624420166015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.01,1.9050432205200196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.01,2.565132713317871
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.16369279623031616
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.21511681079864503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.3109568119049072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.38801279067993166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.4492288112640381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.6004415988922119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.5668416023254395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.6146304130554199
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,0.6227776050567627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,0.5872064113616944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,0.6550144195556641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,0.6414720058441162
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,0.6575488090515137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,0.6885056018829345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,0.7281343936920166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,0.760595178604126
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,0.818175983428955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,0.9211392402648926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.0316479682922364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.241331195831299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,1.3008383750915526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,1.6602048873901367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,2.0078720092773437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,2.6649215698242186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,3.369881439208984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.10739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.12221440076828002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.20436480045318603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.26296319961547854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.26241919994354246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.2653503894805908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.2952192068099976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.27889280319213866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.2856575965881348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.3058495998382568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.29704320430755615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.31687679290771487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.3296704053878784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.3327167987823486
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.3543423891067505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.3815743923187256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.42133121490478515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.4960639953613281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,power_law_1.2,0.5454592227935791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,power_law_1.2,0.7535168170928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,power_law_1.2,0.6936384201049804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,power_law_1.2,0.9079551696777344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,power_law_1.2,1.207084846496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,2,power_law_1.2,0.19133440256118775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,2,power_law_1.2,0.22161920070648194
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,2,power_law_1.2,0.2697664022445679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,2,power_law_1.2,0.36913280487060546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,2,power_law_1.2,0.48197121620178224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,2,power_law_1.2,0.5544960021972656
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,2,power_law_1.2,0.7803967952728271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,2,power_law_1.2,0.8369791984558106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,2,power_law_1.2,0.8148351669311523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,power_law_1.2,1.5483136177062988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,2,power_law_1.2,0.8174976348876953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,2,power_law_1.2,0.8566464424133301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,2,power_law_1.2,0.9148799896240234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,2,power_law_1.2,0.8959872245788574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,2,power_law_1.2,0.8857151985168457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,2,power_law_1.2,0.9661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,2,power_law_1.2,0.973964786529541
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,power_law_1.2,1.8718015670776367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,2,power_law_1.2,0.9837504386901855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,2,power_law_1.2,1.058195209503174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.2,1.1155584335327149
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.2,1.199020767211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.2,1.3451071739196778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.2,1.5086527824401856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.2,1.8134336471557617
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.2,2.377369689941406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.2,2.9124992370605467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.2,4.164723205566406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.2,5.355142211914062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,balanced,0.03925866633653641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,balanced,0.03818133225043615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,balanced,0.04142399877309799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,balanced,0.0421973317861557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,balanced,0.04112533231576284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,balanced,0.04649066428343455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,balanced,0.04554133117198944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,balanced,0.05060266455014547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,balanced,0.05505600074927012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,balanced,0.06577600042025249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,balanced,0.07881600161393483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,balanced,0.09494933485984802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,balanced,0.11990400155385335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,balanced,0.14291733503341675
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.12970240116119386
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.12508800029754638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.12604160308837892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.12680959701538086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.13164160251617432
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.1421504020690918
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.15261440277099608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.16158080101013184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.17787519693374634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.19674880504608155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.2174527883529663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.25867519378662107
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.32332160472869875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,power_law_1.01,0.4107840061187744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,power_law_1.01,0.560268783569336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,power_law_1.01,0.739359998703003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,power_law_1.01,0.09041919708251953
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,power_law_1.01,0.10284160375595093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,power_law_1.01,0.09678080081939697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,power_law_1.01,0.10011520385742187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,power_law_1.01,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,power_law_1.01,0.09826560020446777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,power_law_1.01,0.10439679622650147
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,power_law_1.01,0.09834240078926086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,power_law_1.01,0.10055040121078491
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,power_law_1.01,0.1023360013961792
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,power_law_1.01,0.10307199954986572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,power_law_1.01,0.11659519672393799
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,power_law_1.01,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,power_law_1.01,0.1194111943244934
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,power_law_1.01,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.01,0.14180480241775512
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.01,0.17007999420166015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.01,0.18755199909210205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.01,0.23743999004364014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.01,0.281331205368042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.01,0.38217599391937257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.01,0.47424001693725587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.01,0.7256447792053222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.01,0.912441635131836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,power_law_1.01,0.9149824142456054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,power_law_1.01,1.3951935768127441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,power_law_1.01,2.252467155456543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,power_law_1.01,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,power_law_1.01,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,power_law_1.01,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,power_law_1.01,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,power_law_1.01,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,power_law_1.01,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,power_law_1.01,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,power_law_1.01,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,power_law_1.01,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,power_law_1.01,0.09768319725990296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,power_law_1.01,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.01,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.01,0.13290239572525026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.01,0.1531391978263855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.01,0.21340160369873046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.01,0.24951679706573487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.01,0.3561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.01,0.4210559844970703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.01,0.7362112045288086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.01,0.8538687705993653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.11560959815979004
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.13953280448913574
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.17420159578323363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.20899839401245118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.23102080821990967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.28092160224914553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.2962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.30020480155944823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.3124351978302002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.3263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.33635199069976807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.34766719341278074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.3630847930908203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.367903995513916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.39027841091156007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.4462143898010254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.4863743782043457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,0.5883711814880371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,0.6762432098388672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,0.8631232261657715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,1.0476287841796874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,1.4127231597900392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.2,1.8941503524780274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.2,2.6240255355834963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.2,3.240678405761719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,power_law_1.2,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,power_law_1.2,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,power_law_1.2,0.02837119996547699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,power_law_1.2,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,power_law_1.2,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.2,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.2,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.2,0.13902080059051514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.2,0.17210880517959595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.2,0.2522559881210327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.2,0.3649600028991699
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.2,0.46392321586608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.2,0.6939839839935302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.11027840375900269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.13140480518341063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.19498239755630492
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.2,1.137171173095703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.22087039947509765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,power_law_1.2,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,power_law_1.2,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,power_law_1.2,0.050195199251174924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,power_law_1.2,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,power_law_1.2,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,power_law_1.2,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,power_law_1.2,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,power_law_1.2,0.06339840292930603
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,power_law_1.2,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,power_law_1.2,0.0706816017627716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.33966081142425536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,power_law_1.2,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,power_law_1.2,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,power_law_1.2,0.09686400294303894
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,power_law_1.2,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,power_law_1.2,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,power_law_1.2,0.15060479640960694
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,power_law_1.2,0.1852671980857849
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,power_law_1.2,0.23978879451751708
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,power_law_1.2,0.30379519462585447
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,power_law_1.2,0.4540863990783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.44097280502319336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,power_law_1.2,0.5581183910369873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,power_law_1.2,0.753446388244629
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,balanced,0.032298666735490165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,balanced,0.05306666592756907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,balanced,0.07977599898974101
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,balanced,0.11939733227094014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,balanced,0.1241439978281657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,balanced,0.12271466851234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,balanced,0.1225920021533966
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,balanced,0.1232266624768575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,balanced,0.12432533502578735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.01,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,balanced,0.12427199880282085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,balanced,0.1272266705830892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,balanced,0.12686399618784586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,balanced,0.1297706663608551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,balanced,0.13757866621017456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,balanced,0.13700800140698752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,balanced,0.14300266901652017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,balanced,0.15176000197728476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,balanced,0.16709333658218384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,balanced,0.18550399939219156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.01,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,balanced,0.2023520072301229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,balanced,0.2535253365834554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.01,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,balanced,0.2956800063451131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.01,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,balanced,0.400762677192688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,balanced,0.47499199708302814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,balanced,0.6579413414001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.01,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.01,0.04275839924812317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,balanced,0.8423360188802084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.01,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.01,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.01,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.01,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.01,0.09228159785270691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.01,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.01,0.14228479862213134
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.01,0.17415039539337157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.01,0.23021440505981444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.01,0.27774078845977784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,power_law_1.2,1.09749755859375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,balanced,0.0939359962940216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,balanced,0.14500266313552856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,balanced,0.23741867144902548
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,balanced,0.4385066827138265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,balanced,0.8241653442382812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,balanced,1.605807940165202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,balanced,1.6111733118693035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,balanced,1.5999892552693684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,balanced,1.6023252805074055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,balanced,1.6088107426961262
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,balanced,1.6183360417683919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,balanced,1.6273013750712078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,balanced,1.6353012720743816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,balanced,1.6321172714233398
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,balanced,1.6427466074625652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,balanced,1.6505866050720215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,balanced,1.6672746340433757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,balanced,1.6937920252482097
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,balanced,1.7357652982076008
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,balanced,1.7830026944478352
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,balanced,1.846992015838623
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,balanced,1.9267840385437012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,balanced,2.077237288157145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,balanced,2.302015940348307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,balanced,2.477274735768636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,balanced,3.389568010965983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,balanced,3.7548853556315103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,balanced,0.03173333406448364
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,balanced,0.05235200126965841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,balanced,0.0688266654809316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,balanced,0.08212799827257793
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,balanced,0.07838933169841766
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,balanced,0.07699733475844066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,balanced,0.07789866626262665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,balanced,0.07819733520348866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,balanced,0.07912000020345052
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,balanced,0.07974933087825775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,balanced,0.08331199983755748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,balanced,0.08310933411121368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,balanced,0.09050666292508443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,balanced,0.1172213355700175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,balanced,0.10121599833170573
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,balanced,0.1263200044631958
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,balanced,0.12103999654452006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,balanced,0.16289066274960837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,balanced,0.17559999227523804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,balanced,0.2529760003089905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,balanced,0.2532426714897156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,balanced,0.052383999029795326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,balanced,0.054789334535598755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,balanced,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,balanced,0.09967466195424397
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,balanced,0.11897599697113037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,balanced,0.13476799925168356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,balanced,0.17075733343760172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,balanced,0.20402665932973227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,balanced,0.2930399974187215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,balanced,0.34910400708516437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,balanced,0.04609066744645437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,balanced,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,balanced,0.07702933251857758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,balanced,0.11188266674677531
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,balanced,0.11251733700434367
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,balanced,0.11289067069689433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,balanced,0.11151466766993205
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,balanced,0.1148373285929362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,balanced,0.11591466267903645
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,balanced,0.11788800358772278
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,balanced,0.1184213360150655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,balanced,0.12069867054621379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,balanced,0.1269866625467936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,balanced,0.13116799791653952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,balanced,0.13966932892799377
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,balanced,0.1502240002155304
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,balanced,0.15894400080045065
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,balanced,0.18032532930374146
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,balanced,0.20381865898768106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,balanced,0.2629599968592326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,balanced,0.305461327234904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,balanced,0.421776016553243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,balanced,0.5173759857813517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,balanced,0.7321226596832275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,balanced,0.9173866907755533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.06971520185470581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.08318079710006714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.10828160047531128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.11118079423904419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.11407999992370606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.11874560117721558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.12654080390930175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.12999680042266845
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.13189760446548462
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.1579327940940857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.1725759983062744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.2425407886505127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.3000511884689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.36706559658050536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,power_law_1.01,0.47506561279296877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,power_law_1.01,0.6157887935638428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,power_law_1.01,0.9321663856506348
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,power_law_1.01,1.1384767532348632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.01,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.01,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.01,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.01,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.01,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.01,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.01,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.01,0.06238719820976257
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.01,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.01,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.01,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.01,0.10552959442138672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.01,0.12120319604873657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.01,0.15341440439224244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.01,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.01,0.246560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.01,0.25846400260925295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,balanced,0.06843199829260509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,balanced,0.06912533442179362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,balanced,0.0691893349091212
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,balanced,0.07113066812356313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,balanced,0.07116266588370006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,balanced,0.07072000205516815
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,balanced,0.07181866466999054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,balanced,0.07252266506354015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,balanced,0.08025066554546356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,balanced,0.08707200487454732
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,balanced,0.08855467041333516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,balanced,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,balanced,0.11063999931017558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.1595200002193451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.18012799819310507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.2359679937362671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.27899734179178876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.3930453459421794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.4928640127182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.7016586462656657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.919269323348999
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,balanced,0.057818666100502014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,balanced,0.08302933474381764
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,balanced,0.11413333813349406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,balanced,0.12993066509564719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,balanced,0.1341333289941152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,balanced,0.13397333025932312
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,balanced,0.13659200072288513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,balanced,0.13794133067131042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,balanced,0.1379680037498474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,balanced,0.14199999968210855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,balanced,0.1458133359750112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,power_law_1.2,0.06713600158691406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,power_law_1.2,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,power_law_1.2,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,power_law_1.2,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,power_law_1.2,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,power_law_1.2,0.0961023986339569
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,balanced,0.15083733201026917
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,balanced,0.15762666861216226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,balanced,0.16352533300717673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,balanced,0.1697280009587606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,balanced,0.19526932636896768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,balanced,0.21144000689188638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,balanced,0.2709386746088664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,balanced,0.2898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,balanced,0.41626667976379395
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,balanced,0.4872639973958333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,balanced,0.7136106491088867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,balanced,0.852463960647583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,power_law_1.2,0.09671679735183716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,balanced,1.2651093006134033
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,power_law_1.2,0.09649919867515563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,balanced,1.5926720301310222
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,power_law_1.2,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,power_law_1.2,0.0997439980506897
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,power_law_1.2,0.11079679727554322
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,power_law_1.2,0.11438720226287842
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,power_law_1.2,0.12085119485855103
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,power_law_1.2,0.14523520469665527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,power_law_1.2,0.15259519815444947
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,power_law_1.2,0.1742784023284912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,power_law_1.2,0.2309056043624878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.2,0.28506879806518554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.2,0.37610878944396975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.2,0.44591360092163085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.2,0.6213312149047852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.2,0.7931392192840576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.2,1.1985600471496582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.2,1.5750656127929688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.2,2.3330047607421873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.2,3.448403167724609
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,balanced,0.08052266637484233
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,balanced,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,balanced,0.1716266671816508
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,balanced,0.2858560085296631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,balanced,0.4035573403040568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,balanced,0.4029546578725179
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,balanced,0.40067732334136963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,balanced,0.4045120080312093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,balanced,0.4105386734008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,balanced,0.4171786705652873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,balanced,0.42137598991394043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,balanced,0.42744000752766925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,balanced,0.43939733505249023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,balanced,0.44469332695007324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,balanced,0.4611039956410726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,balanced,0.509002685546875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,balanced,0.5370399951934814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,balanced,0.6114879846572876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,balanced,0.673904021581014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,balanced,0.8162986437479655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,balanced,1.0030133724212646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,balanced,1.2810346285502117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,balanced,1.6864852905273438
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,balanced,2.2630772590637207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,balanced,3.03057066599528
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.09210879802703857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.08216320276260376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.08826879858970642
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.09326080083847046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.09785599708557129
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.11287039518356323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.12729599475860595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.16302720308303834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.17908480167388915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.22889599800109864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.27282559871673584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.3697727918624878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.4421823978424072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.6650623798370361
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.7978367805480957
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,balanced,0.05913066864013672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,balanced,0.08005333443482716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,balanced,0.12356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,balanced,0.1251626710096995
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,balanced,0.12314132849375407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,balanced,0.1255519986152649
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,balanced,0.12637333075205484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,balanced,0.12760000427563986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,balanced,0.1283253331979116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,balanced,0.12774399916330972
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,balanced,0.13149333000183105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,balanced,0.12973333398501077
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,balanced,0.1300373375415802
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,balanced,0.13959999879201254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,balanced,0.15204266707102457
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,balanced,0.15641599893569946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,balanced,0.1597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,balanced,0.19057067235310873
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,balanced,0.1980746587117513
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,balanced,0.24276800950368246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,balanced,0.2767146627108256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,balanced,0.37169067064921063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,balanced,0.46452800432840985
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,balanced,0.6537546714146932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,balanced,0.8387413024902344
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,balanced,1.2249600092569988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,balanced,1.6096320152282715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,power_law_1.01,0.10574079751968384
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,power_law_1.01,0.11913599967956542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,power_law_1.01,0.14036480188369752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,power_law_1.01,0.15911680459976196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,power_law_1.01,0.1758015990257263
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,power_law_1.01,0.20824320316314698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,power_law_1.01,0.2103935956954956
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,power_law_1.01,0.21638400554656984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,power_law_1.01,0.21310720443725586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,power_law_1.01,0.22892799377441406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,power_law_1.01,0.2311039924621582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,balanced,0.05972800155480703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,power_law_1.01,0.23456640243530275
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,power_law_1.01,0.23927040100097657
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,2,power_law_1.01,0.23992960453033446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,2,power_law_1.01,0.2371583938598633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,2,power_law_1.01,0.2534208059310913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,balanced,0.06070399781068166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,2,power_law_1.01,0.2716032028198242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,balanced,0.06557333469390869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,balanced,0.066021333138148
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,balanced,0.07422933479150136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,balanced,0.07761066655317943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,balanced,0.08860266208648682
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.01,0.2808511972427368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,balanced,0.09913067022959392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,balanced,0.1193333367506663
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,balanced,0.1393013298511505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,balanced,0.19548267126083374
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,balanced,0.2339413364728292
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,balanced,0.33074132601420086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,balanced,0.4283999999364217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,balanced,0.6301440000534058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,balanced,0.8278240362803141
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.01,0.32065279483795167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.01,0.3728127956390381
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.01,0.45713281631469727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.01,0.5520448207855224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.01,0.721721601486206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.01,0.9368767738342285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.01,1.336134433746338
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.01,1.7209663391113281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,balanced,0.04734399914741516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,balanced,0.053674668073654175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,balanced,0.06628799935181935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,balanced,0.08298666775226593
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,balanced,0.09702933828035991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,balanced,0.11969600121180217
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,balanced,0.13920533657073975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,balanced,0.17744000752766928
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,balanced,0.21572266022364298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,balanced,0.29214932521184284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,balanced,0.36549333731333417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,balanced,0.5231893459955851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,balanced,0.6694560050964355
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.12718720436096193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.1387519955635071
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.17391999959945678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.23036160469055175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.3018239974975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.3752448081970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.5657472133636474
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,0.7001728057861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,balanced,0.04727466901143392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,balanced,0.04832000037034353
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,balanced,0.05374933282534281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,balanced,0.08255466818809509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,balanced,0.1053013304869334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,balanced,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,balanced,0.15652799606323242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,balanced,0.19126399358113608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,balanced,0.25729600588480633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,balanced,0.3291200002034505
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,power_law_1.2,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,power_law_1.2,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,power_law_1.2,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,power_law_1.2,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,power_law_1.2,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,power_law_1.2,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,power_law_1.2,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,power_law_1.2,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,power_law_1.2,0.11045119762420655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,power_law_1.2,0.11485439538955688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,power_law_1.2,0.11831680536270142
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,power_law_1.2,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,power_law_1.2,0.13742719888687133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,power_law_1.2,0.146560001373291
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,power_law_1.2,0.15276800394058226
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,power_law_1.2,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,power_law_1.2,0.17006080150604247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,power_law_1.2,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,power_law_1.2,0.2299583911895752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,power_law_1.2,0.2674495935440063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,power_law_1.2,0.1051584005355835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,power_law_1.2,0.35344638824462893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,power_law_1.2,0.12103040218353271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,power_law_1.2,0.4258431911468506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,power_law_1.2,0.14090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,power_law_1.2,0.5892608165740967
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,power_law_1.2,0.1741055965423584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,power_law_1.2,0.7572927951812745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,power_law_1.2,0.18001279830932618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,power_law_1.2,1.1382335662841796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,power_law_1.2,0.18235520124435425
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,power_law_1.2,0.19032319784164428
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,power_law_1.2,1.4222975730895997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,power_law_1.2,0.19782400131225586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,power_law_1.2,0.20332159996032714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,power_law_1.2,0.20591359138488768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,power_law_1.2,0.2185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,power_law_1.2,0.22483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,power_law_1.2,0.23394560813903809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,power_law_1.2,0.2516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,power_law_1.2,0.2870975971221924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.2,0.3184767961502075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.2,0.38625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.2,0.45269122123718264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.2,0.5879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.2,0.7146048069000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.2,0.9883904457092285
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.2,1.2527551651000977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.2,1.780998420715332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.2,2.309369659423828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,power_law_1.01,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,power_law_1.01,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,power_law_1.01,0.055206400156021115
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,power_law_1.01,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,power_law_1.01,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,power_law_1.01,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,power_law_1.01,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,power_law_1.01,0.06468480229377746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,power_law_1.01,0.07850239872932434
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,power_law_1.01,0.0831167995929718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,power_law_1.01,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.01,0.11349760293960572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.01,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.01,0.15126399993896483
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.01,0.19644800424575806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.01,0.22990078926086427
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.01,0.3210752010345459
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.01,0.40769281387329104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.01,0.5637375831604003
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.01,0.7940671920776368
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,power_law_1.2,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,power_law_1.2,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,power_law_1.2,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,power_law_1.2,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,power_law_1.2,0.10974080562591552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,power_law_1.2,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,power_law_1.2,0.12854399681091308
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,power_law_1.2,0.13930879831314086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,power_law_1.2,0.14632960557937622
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,power_law_1.2,0.15762560367584227
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,power_law_1.2,0.1651136040687561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,power_law_1.2,0.1778880000114441
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,power_law_1.2,0.19950720071792602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,power_law_1.2,0.23451519012451172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,power_law_1.2,0.24133119583129883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,power_law_1.2,0.2986687898635864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,power_law_1.2,0.3535295963287354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,power_law_1.2,0.47280001640319824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,power_law_1.2,0.570963191986084
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,power_law_1.2,0.80764799118042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,power_law_1.2,1.0028544425964356
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,power_law_1.2,1.4592767715454102
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,power_law_1.2,1.8932479858398437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,balanced,0.056458666920661926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,balanced,0.10006933410962422
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,balanced,0.10227732857068379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,balanced,0.10181867082913716
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,balanced,0.10109866658846538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,balanced,0.10434133807818095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,balanced,0.10293333729108174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,balanced,0.10393599669138591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,balanced,0.10496532917022705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,balanced,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,balanced,0.10870400071144104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,balanced,0.11217066645622253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,balanced,0.11288533608118693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,balanced,0.116949329773585
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,balanced,0.1242026686668396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,balanced,0.13182399670283
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,balanced,0.14800000190734863
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,balanced,0.16369066635767618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,balanced,0.21500800053278604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,balanced,0.2526719967524211
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,balanced,0.35994664827982586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,balanced,0.460970679918925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,balanced,0.6701066493988037
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,balanced,0.8828907012939453
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.05961599946022034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.08611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.07868160009384155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.10248960256576538
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.09819520115852357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.1587391972541809
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.22232959270477295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.2419071912765503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.3309056043624878
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,power_law_1.2,0.4259456157684326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,power_law_1.2,0.6248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,power_law_1.2,0.7819519996643066
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,power_law_1.2,1.1668288230895996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,balanced,0.036464000741640724
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,balanced,0.04433600107828776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,balanced,0.05633600056171417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,balanced,0.06865066786607106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,balanced,0.10528533657391866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,balanced,0.1291253368059794
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,power_law_1.2,1.4160767555236817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,balanced,0.03806933263937632
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,balanced,0.0400693342089653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,balanced,0.053397332628568016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,balanced,0.08275733391443889
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,balanced,0.08572799960772197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,balanced,0.08499733606974284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,balanced,0.0904266635576884
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,balanced,0.09046399593353271
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,balanced,0.09281599521636963
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,balanced,0.09799999992052714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,balanced,0.10353599985440572
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,balanced,0.10916266838709514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,balanced,0.11753066380818684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,balanced,0.13580800096193948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,balanced,0.15218666195869446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,balanced,0.20670400063196817
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,balanced,0.2358986735343933
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,balanced,0.35025068124135333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,balanced,0.403711994489034
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,balanced,0.5748533407847086
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,balanced,0.7250613371531168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,balanced,1.084330638249715
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,balanced,1.4048800468444824
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.12066559791564942
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.1307904005050659
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.13054720163345337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.1368064045906067
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.1295232057571411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.1450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.13857280015945433
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.1534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.14832639694213867
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.15688960552215575
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.1602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.16469759941101075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.1814144015312195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.19415040016174318
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.23853440284729005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.2542912006378174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.30872321128845215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.3641664028167725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,power_law_1.2,0.4558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,power_law_1.2,0.5647232055664062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,power_law_1.2,0.7749440193176269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,power_law_1.2,1.031167984008789
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,balanced,0.025285333395004272
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,balanced,0.023418667415777843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,balanced,0.02515200028816859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,balanced,0.025077333052953083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,balanced,0.025050667424996693
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,balanced,0.0272533322374026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,balanced,0.027301333844661713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,balanced,0.02920000006755193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,balanced,0.030986666679382324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,balanced,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,balanced,0.055530667304992676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,balanced,0.08012266457080841
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,balanced,0.09148266911506653
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,balanced,0.031023999055226643
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,balanced,0.033226666351159416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,balanced,0.031301334500312805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,balanced,0.03319466610749563
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,balanced,0.08251733581225078
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,balanced,0.12138133247693379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,balanced,0.15173332889874777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,balanced,0.21103467543919882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,balanced,0.25378666321436566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,power_law_1.2,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,power_law_1.2,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,power_law_1.2,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,power_law_1.2,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,power_law_1.2,0.049830400943756105
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,power_law_1.2,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,power_law_1.2,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,power_law_1.2,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,power_law_1.2,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,power_law_1.2,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,power_law_1.2,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,power_law_1.2,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,power_law_1.2,0.14714239835739135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,power_law_1.2,0.17876479625701905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,power_law_1.2,0.22678399085998535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,power_law_1.2,0.2944576025009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,power_law_1.2,0.43785600662231444
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,power_law_1.2,0.5182271957397461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.11621760129928589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.15251200199127196
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.1416640043258667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.1584447979927063
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.16054400205612182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.1547711968421936
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.15745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.1668544054031372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.16727039813995362
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.16949119567871093
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.1944640040397644
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.20592000484466552
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.22213120460510255
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.2593600034713745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.2776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.3360512018203735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.41675519943237305
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.5201087951660156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.6446400165557862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.8790719985961915
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.12423677444458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.2,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.2,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.2,0.08762879967689514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.2,0.0905344009399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.2,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.2,0.09519360065460206
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.2,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.2,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.2,0.1047935962677002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.2,0.11083519458770752
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.2,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.2,0.1271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.2,0.1439743995666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.2,0.1661631941795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.2,0.19881600141525269
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.2,0.237555193901062
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.2,0.3014271974563599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.2,0.33540480136871337
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.2,0.46459522247314455
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.2,0.5793983936309814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.2,0.8349568367004394
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.2,1.134438419342041
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,power_law_1.01,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,power_law_1.01,0.04047360122203827
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,power_law_1.01,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,power_law_1.01,0.04919039905071258
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,power_law_1.01,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,power_law_1.01,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,power_law_1.01,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,power_law_1.01,0.07296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,power_law_1.01,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,power_law_1.01,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,power_law_1.01,0.12649600505828856
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,power_law_1.01,0.1769536018371582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,power_law_1.01,0.21823360919952392
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,power_law_1.01,0.32914559841156005
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,power_law_1.01,0.39246718883514403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,power_law_1.01,0.5905216217041016
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,power_law_1.01,0.7556159973144532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,power_law_1.2,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,power_law_1.2,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,power_law_1.2,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,power_law_1.2,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,power_law_1.2,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,power_law_1.2,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,power_law_1.2,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,power_law_1.2,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,power_law_1.2,0.10696320533752442
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,power_law_1.2,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,power_law_1.2,0.1601088047027588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,power_law_1.2,0.20093441009521484
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,power_law_1.2,0.25029759407043456
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,power_law_1.2,0.3670975923538208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,power_law_1.2,0.4416639804840088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,power_law_1.2,0.6410624027252197
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,power_law_1.2,0.8128000259399414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.05601919889450073
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.11127680540084839
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.14455679655075074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.17301119565963746
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.2205888032913208
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.30458879470825195
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.3364736080169678
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.5042304039001465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.6371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.08412799835205079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.12904959917068481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.1292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.13163520097732545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.13241599798202514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.13468159437179567
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.14094719886779786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.14343680143356324
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.14631680250167847
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.15854719877243043
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.15893759727478027
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.17329920530319215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.19166719913482666
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.2148736000061035
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.2684992074966431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.2762687921524048
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.35738239288330076
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.3958143949508667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.5742720127105713
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.6918655872344971
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,1.0025279998779297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.3462656021118165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,balanced,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,balanced,0.09141866366068523
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,balanced,0.09621866544087727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,balanced,0.09505599737167358
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,balanced,0.09775466720263164
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,balanced,0.09650133053461711
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,balanced,0.09731200337409973
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,balanced,0.09796266754468282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,balanced,0.09877333045005798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,balanced,0.09802666306495667
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,balanced,0.09988266229629517
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,balanced,0.1015786627928416
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,balanced,0.11155733466148376
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,balanced,0.11514666676521301
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,balanced,0.15451199809710184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,balanced,0.14689600467681885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,balanced,0.19631467262903848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,balanced,0.22515199581782022
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,balanced,0.304149329662323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,balanced,0.3995680014292399
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,balanced,0.5617813269297282
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,balanced,0.7367146809895834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,balanced,1.0941759745279949
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,balanced,1.4247946739196777
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,balanced,0.06594666838645935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,balanced,0.16865599155426025
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,balanced,0.2950399915377299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,balanced,0.3890399932861328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,balanced,0.3922079801559448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,balanced,0.39137065410614014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,balanced,0.39452266693115234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,balanced,0.39559467633565265
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,balanced,0.39671464761098224
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,balanced,0.40112535158793133
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,balanced,0.4022666613260905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,balanced,0.4063520034154256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,balanced,0.4131786823272705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,balanced,0.4152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,balanced,0.423093318939209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,balanced,0.43585066000620526
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,balanced,0.4474613269170125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,balanced,0.5401440064112345
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,balanced,0.5068480173746744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,balanced,0.7501280307769775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,balanced,0.6293066740036011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,balanced,0.9395360151926676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,balanced,0.9400160312652588
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,balanced,1.3253653049468994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,balanced,1.4469173749287922
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,power_law_1.01,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,power_law_1.01,0.09434239864349366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,power_law_1.01,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,power_law_1.01,0.11447039842605591
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,power_law_1.01,0.11644159555435181
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,power_law_1.01,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,power_law_1.01,0.1436352014541626
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,power_law_1.01,0.14365439414978026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,power_law_1.01,0.14332159757614135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,power_law_1.01,0.14609919786453246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,power_law_1.01,0.14839040040969848
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,power_law_1.01,0.1496448040008545
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,power_law_1.01,0.15389440059661866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,power_law_1.01,0.15605759620666504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,power_law_1.01,0.16021759510040284
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,power_law_1.01,0.1641343951225281
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,power_law_1.01,0.17032959461212158
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,power_law_1.01,0.19017599821090697
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.01,0.2124095916748047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.01,0.2553152084350586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.01,0.3014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.01,0.40170879364013673
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.01,0.5062784194946289
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.01,0.6971136093139648
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.01,0.8815296173095704
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.01,1.2713151931762696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.01,1.6486591339111327
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,balanced,0.05040533343950907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,balanced,0.057717333237330117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,balanced,0.05981333553791046
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,balanced,0.07464000085989635
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,balanced,0.08273600041866302
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,balanced,0.09901866316795349
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,balanced,0.11521599690119426
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,balanced,0.15638400117556253
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,balanced,0.18360533316930136
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,balanced,0.24833599726359049
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,balanced,0.3097599943478902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,power_law_1.01,0.02476159930229187
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,power_law_1.01,0.023347200453281404
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,power_law_1.01,0.02389120012521744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,power_law_1.01,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,power_law_1.01,0.02476799935102463
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,power_law_1.01,0.026054400205612182
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,power_law_1.01,0.026086398959159852
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,power_law_1.01,0.027897599339485168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,power_law_1.01,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,power_law_1.01,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,balanced,0.1390880048274994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,balanced,0.14035733540852866
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,balanced,0.13896532853444418
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,balanced,0.13994133472442627
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,balanced,0.1426346699396769
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,balanced,0.14406933387120566
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,balanced,0.1418293317159017
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,balanced,0.14748799800872803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,balanced,0.14786666631698608
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,balanced,0.15037332971890768
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,balanced,0.15286399920781454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,balanced,0.16872000694274902
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,balanced,0.17171200116475424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,balanced,0.18038400014241537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,balanced,0.215664009253184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,balanced,0.22907199462254843
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,balanced,0.28643733263015747
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,balanced,0.3233013351758321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,balanced,0.43777068456013996
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,power_law_1.01,0.030220800638198854
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,power_law_1.01,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,balanced,0.543173352877299
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,balanced,0.778602679570516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,balanced,0.9838506380716959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,balanced,1.4580693244934082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,balanced,1.9050347010294597
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.15194239616394042
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.1846336007118225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,balanced,0.023039999107519787
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,balanced,0.025125332176685333
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,balanced,0.02521066615978877
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,balanced,0.022991999983787537
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,balanced,0.025077333052953083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,balanced,0.025386666258176167
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,balanced,0.025029333929220837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,balanced,0.02517866591612498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,balanced,0.02516266703605652
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,balanced,0.02720000098148982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,balanced,0.025461333493391674
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,balanced,0.027082666754722595
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,balanced,0.030847998956839245
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,balanced,0.033157333731651306
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,balanced,0.03934400031963984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,balanced,0.057573333382606506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,power_law_1.01,0.07596160173416137
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,power_law_1.01,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,power_law_1.01,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,power_law_1.01,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,power_law_1.01,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,power_law_1.01,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,power_law_1.01,0.07130879759788514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,power_law_1.01,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,power_law_1.01,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,power_law_1.01,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,power_law_1.01,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,power_law_1.01,0.10677759647369385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,power_law_1.01,0.11412479877471923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,power_law_1.01,0.1441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,power_law_1.01,0.15798399448394776
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,power_law_1.01,0.20424959659576417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,power_law_1.01,0.2507008075714111
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,power_law_1.01,0.3338943958282471
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,power_law_1.01,0.44208641052246095
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,power_law_1.01,0.6394432067871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,power_law_1.01,0.7720895767211914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.01,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.01,0.20151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.01,0.32511999607086184
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.01,0.5349631786346436
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.01,0.7576320171356201
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.01,1.215609645843506
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.01,1.3073216438293458
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.01,1.3713727951049806
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.01,1.388211154937744
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.01,1.419871997833252
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.01,1.4642687797546388
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.01,1.493068790435791
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.01,1.5074624061584472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.01,1.574284839630127
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.01,1.611065673828125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.01,1.664473533630371
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.01,1.7397567749023437
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.01,1.9635263442993165
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.01,1.9183679580688477
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.01,2.1779584884643555
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.01,2.2399551391601564
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.01,2.7741504669189454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,power_law_1.01,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.01,2.9885120391845703
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,power_law_1.01,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.01,3.6645503997802735
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,power_law_1.01,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.01,4.374118423461914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,power_law_1.01,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,power_law_1.01,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,power_law_1.01,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.01,6.0726463317871096
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,power_law_1.01,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,power_law_1.01,0.11118719577789307
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,power_law_1.01,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,power_law_1.01,0.17774720191955568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,power_law_1.01,0.23486719131469727
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,power_law_1.01,0.2761663913726807
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.01,6.946240234375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,power_law_1.01,0.3905663967132568
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,power_law_1.01,0.4947904109954834
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,power_law_1.01,0.6985856056213379
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,power_law_1.01,0.9465791702270507
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.2,0.1935807943344116
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.2,0.2756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.2,0.4277184009552002
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.2,0.5916160106658935
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.2,0.7628032207489014
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.2,1.2630911827087403
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.2,1.2930047988891602
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.2,1.2989631652832032
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.2,1.3414527893066406
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.2,1.400607967376709
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.2,1.4567680358886719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.2,1.5173376083374024
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.2,1.519148826599121
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.2,1.6042943954467774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.2,1.6225215911865234
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.2,1.6674816131591796
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.2,1.7301759719848633
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.2,1.9043071746826172
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.2,1.9746559143066407
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.2,2.2688255310058594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.2,2.3270912170410156
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.2,2.843596839904785
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.2,2.8979135513305665
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.2,3.551833724975586
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.2,4.640697479248047
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.2,6.1941886901855465
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.2,6.7704833984375
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,balanced,0.05844266712665558
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,balanced,0.06790400048096974
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,balanced,0.08633599678675334
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,balanced,0.11141866445541382
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,balanced,0.11657599608103435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,balanced,0.1163200040658315
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,balanced,0.11564266681671143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,balanced,0.11877333124478658
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,balanced,0.1193386713663737
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,balanced,0.12051199873288472
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,balanced,0.12198932965596516
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,balanced,0.12473066647847493
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,balanced,0.1283146639664968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,balanced,0.12983999649683634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,balanced,0.13083199659983316
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,balanced,0.1404906709988912
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,balanced,0.1508746643861135
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,balanced,0.1618613302707672
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,balanced,0.1917653282483419
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,balanced,0.22430400053660074
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,balanced,0.32094399134318036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,balanced,0.38203732172648114
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,balanced,0.5475466648737589
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,balanced,0.7026080290476481
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,balanced,1.0217706362406414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,balanced,1.3538293838500977
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,power_law_1.01,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,power_law_1.01,0.026579201221466064
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,power_law_1.01,0.02656640112400055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,power_law_1.01,0.02871679961681366
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,power_law_1.01,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,power_law_1.01,0.029241600632667543
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,power_law_1.01,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,power_law_1.01,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,power_law_1.01,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,power_law_1.01,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,power_law_1.01,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,power_law_1.01,0.06929280161857605
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,power_law_1.01,0.09662719964981079
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,power_law_1.01,0.12231680154800414
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,power_law_1.01,0.1835584044456482
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,power_law_1.01,0.2323456048965454
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,balanced,0.08050133287906647
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,balanced,0.10543466607729594
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,balanced,0.15971199671427408
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,balanced,0.26525332530339557
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,balanced,0.32078399260838825
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,balanced,0.3242560029029846
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,balanced,0.3298719922701518
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,balanced,0.331498662630717
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,balanced,0.32958932717641193
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,balanced,0.3378239870071411
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,balanced,0.3422400156656901
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,balanced,0.3449920018513997
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,balanced,0.3521866798400879
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,balanced,0.36271464824676514
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,balanced,0.372874657313029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,balanced,0.3885279893875122
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,balanced,0.4203146696090698
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,balanced,0.4522720177968343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,balanced,0.5800053278605143
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,balanced,0.6151520013809204
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,balanced,0.9192480246225992
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,balanced,0.9314560095469157
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,balanced,1.3824426333109539
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,balanced,1.6223519643147786
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,balanced,2.4444053967793784
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,balanced,3.0493545532226562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,balanced,0.051445335149765015
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,balanced,0.10552000006039937
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,balanced,0.1560479998588562
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,balanced,0.2541439930597941
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,balanced,0.3598666588465373
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,balanced,0.355840007464091
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,balanced,0.3588800032933553
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,balanced,0.3647093375523885
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,balanced,0.3712746699651082
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,balanced,0.37616535027821857
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,balanced,0.37854401270548504
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,balanced,0.38548266887664795
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,balanced,0.3904533386230469
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,balanced,0.39610668023427326
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,balanced,0.4119946559270223
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,balanced,0.4354666471481323
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,balanced,0.4604373375574748
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,balanced,0.5119040012359619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,balanced,0.556821346282959
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,balanced,0.6598933140436808
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,balanced,0.784010648727417
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,balanced,0.9856533209482828
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,balanced,1.2426559925079346
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,balanced,1.6336906750996907
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,balanced,2.1709493001302085
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,balanced,0.07123200098673503
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,balanced,0.08942400415738423
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,balanced,0.12622400124867758
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,balanced,0.1863200068473816
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,balanced,0.313920001188914
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,balanced,0.31465067466100055
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,balanced,0.3194506764411926
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,balanced,0.31726400057474774
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,balanced,0.31970133384068805
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,balanced,0.3235039909680684
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,balanced,0.3260693351427714
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,balanced,0.33099732796351117
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,balanced,0.333568016688029
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,balanced,0.3468159834543864
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3513386646906535
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,balanced,0.3619946638743083
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,balanced,0.3899306853612264
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.4165866772333781
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.4666186571121216
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.5108000040054321
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.593450665473938
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.6769759654998779
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.9356799920399984
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,balanced,1.0909813245137532
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.5697760581970215
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.9495147069295247
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,balanced,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,balanced,0.06241066753864288
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,balanced,0.0706826647122701
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,balanced,0.07241066793600719
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,balanced,0.07263466715812683
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,balanced,0.0744053324063619
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,balanced,0.07434666653474171
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,balanced,0.07681066791216533
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,balanced,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,balanced,0.07850666840871175
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,balanced,0.08453866839408875
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,balanced,0.08663466572761536
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,balanced,0.09086400270462036
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.10911466677983601
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.13025066256523132
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.15523200233777365
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.18655999501546225
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.21602133909861246
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.3019253412882487
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.36369065443674725
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.5214986801147461
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.6696266333262125
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.2,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.2,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.2,0.14088319540023803
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.2,0.1818176031112671
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.2,0.23425281047821045
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.2,0.3620032072067261
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.2,0.38753280639648435
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.2,0.4005824089050293
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.2,0.38190720081329343
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.2,0.4090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.2,0.4155712127685547
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.2,0.4304384231567383
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.2,0.4375296115875244
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.2,0.4540095806121826
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.2,0.474886417388916
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.2,0.4863232135772705
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.2,0.5121407985687256
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.2,0.5824063777923584
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.2,0.6096255779266357
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.2,0.7360832214355468
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.2,0.7920447826385498
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.2,1.0129216194152832
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.2,1.0755392074584962
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.2,1.3879039764404297
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.2,1.808799934387207
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.2,2.672166442871094
TRTLLM,1.2.0rc5,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.2,3.076243209838867
